Overview
The lightweight alternative to Coolify, Dokploy and CapRover
One binary that idles at about 200 MB, with metrics, log search, IAM, audit, previews, rollback and 580+ templates in a single free edition.
- Control plane idle RAM, measured
- ~200 MB
- Capabilities compared, each sourced
- 38
- Templates in the catalog
- 580+
- One edition, no license key
- Apache 2.0
What changes when you switch
Nine platforms people move from, and what Levelrail does better than each, row by row from their own docs.
-
Coolify alternative
Self-hostedAn agent dials out over mTLS to the Docker Engine API instead of SSH from the control plane.
8rows where Levelrail says yes and Coolify says partly or no
-
Dokploy alternative
Self-hostedPlain Docker with a reconciler and a small agent per node, with no Swarm or overlay network.
9rows where Levelrail says yes and Dokploy says partly or no
-
CapRover alternative
Self-hostedPlain Docker containers with a level-triggered reconciler instead of Swarm services.
6rows where Levelrail says yes and CapRover says partly or no
-
Dokku alternative
Self-hostedA dashboard with deploy history and one-click rollback to a pinned previous image in the free edition. Dokku Pro adds a paid web UI.
9rows where Levelrail says yes and Dokku says partly or no
-
Kamal alternative
Self-hostedDesired state lives in a database, with a reconciler recording a status reason after every pass.
No stated minimum
9rows where Levelrail says yes and Kamal says partly or no
-
Vercel alternative
HostedRuns on Linux servers you own, so secrets, metrics and logs stay on your nodes.
Their cloud, no server to size
6rows where Levelrail says yes and Vercel says partly or no
-
Heroku alternative
HostedRuns on Linux boxes you own, with no per-dyno or per-app fee.
Their cloud, no server to size
10rows where Levelrail says yes and Heroku says partly or no
-
Railway alternative
HostedRuns on your Linux servers, so pricing is your own server cost.
Their cloud, no server to size
7rows where Levelrail says yes and Railway says partly or no
-
Cloud Run alternative
HostedA fixed-price server you run, so cost does not move with traffic.
Their cloud, no server to size
10rows where Levelrail says yes and Cloud Run says partly or no
As of 2026-11-18. Statements about other projects come from their own public documentation. RAM bars compare the minimum each project’s docs state with Levelrail's measured idle footprint.
Nine alternatives, 38 rows
Here is how Levelrail stacks up against nine other platforms, row by row, from their own docs. Point at a cell for its note and source, or switch on Notes. A question mark means their docs do not say.
| Capability | Levelrail Self-hosted | Coolify Self-hosted | Dokploy Self-hosted | CapRover Self-hosted | Dokku Self-hosted | Kamal Self-hosted | Vercel Hosted | Heroku Hosted | Railway Hosted | Cloud Run Hosted |
|---|---|---|---|---|---|---|---|---|---|---|
| Documented answersyes, active or partly, no, unstated | 23 / 15 / 0 / 0 23 yes, 15 active, 0 no, 0 not stated, of 38 | 17 / 10 / 4 / 7 17 yes, 10 partly, 4 no, 7 not stated, of 38 | 16 / 8 / 2 / 12 16 yes, 8 partly, 2 no, 12 not stated, of 38 | 7 / 11 / 2 / 18 7 yes, 11 partly, 2 no, 18 not stated, of 38 | 4 / 17 / 5 / 11 4 yes, 17 partly, 5 no, 11 not stated, of 37 | 4 / 12 / 5 / 14 4 yes, 12 partly, 5 no, 14 not stated, of 35 | 12 / 11 / 1 / 8 12 yes, 11 partly, 1 no, 8 not stated, of 32 | 6 / 19 / 1 / 7 6 yes, 19 partly, 1 no, 7 not stated, of 33 | 17 / 11 / 2 / 3 17 yes, 11 partly, 2 no, 3 not stated, of 33 | 11 / 15 / 0 / 7 11 yes, 15 partly, 0 no, 7 not stated, of 33 |
| Architecture | ||||||||||
| Yes. Agents dial out over mTLS; no inbound ports | No. SSH from the control plane to each server | No. Connects to remote servers over SSH keys | No. Nodes added with an SSH key from the dashboard | No. Commands run on the host, over SSH or git push | No. SSH from your machine or CI to each server | Not applicable. | Not applicable. | Not applicable. | Not applicable. | |
No answer in their docs or not applicable: Vercel, Heroku, Railway, Cloud Run. | ||||||||||
| Yes. Desired state converged, status condition with reason per pass | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | No. Event-driven trigger pipeline; no reconcile loop described | No. None: it runs Docker commands per deploy | Not applicable. | Not applicable. | Not applicable. | Not applicable. | |
No answer in their docs or not applicable: Coolify, Dokploy, CapRover, Vercel, Heroku, Railway, Cloud Run. | ||||||||||
| Yes. One control plane binary, plus one agent per extra node | No. Three base containers: app, PostgreSQL and Redis | No. Three parts: Next.js app, PostgreSQL and Traefik | Not stated in their docs. | Not applicable. | Not applicable. | Not applicable. | Not applicable. | Not applicable. | Not applicable. | |
No answer in their docs or not applicable: CapRover, Dokku, Kamal, Vercel, Heroku, Railway, Cloud Run. | ||||||||||
| Observability | ||||||||||
| Yes. Node-local metrics at 15s and full-text log search | Partly. Sentinel historical CPU, memory, disk, network and load charts; log drains to Axiom, New Relic, CloudWatch (new in v4.6.0), Fluent Bit; docs say log queries and dashboards need an external system | Partly. CPU, memory, disk, network metrics with threshold alerts, real-time logs, and v0.31.0 Vector shipping of logs and metrics to external providers; log search not stated | Partly. Optional NetData for CPU, memory, disk, network; app and build logs viewable, no search documented; v1.16.0 adds runtime health, task history and disk alerts | No. Add your own stack | No. Bring your own | Partly. Observability on all plans; Observability Plus ($1.20 per 1M events) adds retention and more metrics; runtime log retention on Enterprise is 3 days, 30 with Plus; CLI log query and request tracing | Partly. Dyno metrics and events shown; log search not documented | Yes. Dashboard CPU, memory, network, disk metrics; log search | Yes. Automatic Cloud Monitoring metrics; logs in Cloud Logging | |
| ||||||||||
| Yes. Crashloop detection surfaces the last log lines | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Partly. Restart policy restarts failing containers; no crashloop detection | Not stated in their docs. | Not stated in their docs. | Partly. Error codes explain crashes in logs; no automatic diagnosis documented | Partly. Default On Failure restart policy restarts up to 10 times; Free and Trial capped at 10 and cannot use Always; paid plans have no restart limit. No automatic diagnosis | Partly. Troubleshooting guide; logs checked manually, no automatic crashloop view | |
No answer in their docs or not applicable: Coolify, Dokploy, CapRover, Kamal, Vercel. | ||||||||||
| Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | |
No answer in their docs or not applicable: Coolify, Dokploy, CapRover, Dokku, Kamal, Vercel, Heroku, Railway, Cloud Run. | ||||||||||
| Deploys | ||||||||||
| Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta | Yes. Redeploys a retained older image; retention is configurable | Yes. Registry-based rollback to any previous deployment image (needs a configured registry); Swarm automatic rollback on failed health check | Yes. Version history (up to 50 saved versions); image rollback does not restore env var values | Partly. Failed image deploys restore the previous image reference; old images are cleaned up unless skip-cleanup is set, and tags:deploy can redeploy a tagged image manually; no one-command rollback | Partly. Rollback to chosen version; old containers pruned after 3 days | Yes. Instant Rollback repoints production domains to an earlier production deployment with no rebuild. Hobby can only roll back to the immediately previous deployment; Pro and Enterprise can pick any eligible one. After a rollback auto-assignment of production domains is turned off until you promote a deployment | Yes. heroku rollback creates a release reusing the earlier slug or image | Partly. Dashboard rollback to prior image; images kept 24 to 360 hours | Yes. Traffic rolls back to a previous revision immediately | |
| ||||||||||
| Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta | Partly. Starts replacement first; docs say no zero-downtime guarantee | Partly. Swarm health check plus update config with automatic rollback on failure is documented (needs a /health endpoint and curl in the image) | Partly. Stateless apps default to start-first with optional health check; apps with mounted data default to stop-first; docs do not promise interruption-free updates | Yes. New containers start, proxy switches, old retired after delay | Yes. Zero-downtime cutover with kamal-proxy | Not stated in their docs. | Partly. Preboot is opt-in, standard and performance dynos; not guaranteed | Partly. Healthcheck gates cutover; volume services see brief downtime | Partly. Gradual rollouts and traffic splits; health gating of cutover not stated | |
No answer in their docs or not applicable: Vercel. | ||||||||||
| Yes. Stable on GitHub, beta on GitLab and Bitbucket | Yes. Per pull request previews; full UI needs a GitHub App | Partly. Preview per pull request, documented for GitHub only | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Yes. Preview deployment with unique branch and commit URLs for every non-production push, PR (GitHub, GitLab, Bitbucket) or CLI deploy. Protected by Vercel Authentication by default; password protection is $20 per project per month on Pro; custom preview suffix $100 per month on Pro | Yes. Review apps per GitHub pull request; GitHub only | Yes. Environments per PR | Partly. Revision tags give separate URLs without traffic; no PR previews | |
No answer in their docs or not applicable: CapRover, Dokku, Kamal. | ||||||||||
| Active. Pre-built images from the dashboard; git deploys can build | Yes. Deploys docker-compose.yml from a git repository or pasted file | Yes. Compose deploys from Git or pasted file; build unsupported in Stack mode | Partly. Experimental import of a Compose subset (image, environment, ports, volumes, depends_on, command); build, networks, secrets, deploy ignored | No. Deployment methods listed are git push, Dockerfile, Docker image and buildpacks; Compose files are not a documented deploy input | Not stated in their docs. | No. Compose file is not read ('Vercel does not read any of it'). Services are declared in vercel.json (Vercel Services and container images are both in beta on all plans); migration is manual and some patterns (bind mounts, depends_on ordering, privileged containers) have no equivalent | No. Docker Compose is mentioned for local development only; deploy options are container:push and release, or heroku.yml builds (heroku.yml Docker builds are unsupported on Fir) | Partly. Imports Compose services; docker-compose.yml not run directly | Partly. gcloud run compose up deploys a Compose file as one multi container service (supported subset of fields, secrets map to Secret Manager, volumes to Cloud Storage, max instances 1 by default); Google recommends Terraform for production | |
No answer in their docs or not applicable: Kamal. | ||||||||||
| Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta | Yes. Push webhooks for GitHub, GitLab and Bitbucket sources | Yes. Auto deploy on push for GitHub, GitLab, Bitbucket and Gitea | Partly. Generic push webhook documented with GitHub steps; other providers via their own webhook screens or CI | Partly. git push to deploy branch; git:sync from any cloneable repo; GitHub push webhook deploys only in paid Dokku Pro 1.4.0; no GitLab or Bitbucket webhooks documented | Not stated in their docs. | Yes. Auto deploys on push from GitHub, GitLab and Bitbucket | Partly. GitHub automatic deploys (GitHub Enterprise Server and Cloud for Enterprise customers); GitLab and Bitbucket go through CI with dpl, git hooks, the Platform API or Terraform | Partly. Autodeploy from GitHub only; GitLab and Bitbucket not documented | Yes. Developer Connect covers GitHub, GitLab, Bitbucket; Cloud Build covers GitHub | |
No answer in their docs or not applicable: Kamal. | ||||||||||
| Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected | Yes. Nixpacks and Railpack detect and build without a Dockerfile | Yes. Nixpacks is the default; Railpack and buildpacks also offered | Partly. Dockerfile, prebuilt image, inline Dockerfile or built-in language templateId | Partly. Heroku buildpacks by default; pack and Railpack need config files | Partly. Dockerfile required by default; buildpacks only when configured explicitly | Yes. Framework auto-detection builds with no Dockerfile needed | Yes. Buildpacks auto-detect language and build source with no Dockerfile | Yes. Railpack detects stack without Dockerfile; Dockerfile takes priority | Yes. gcloud run deploy --source uses a Dockerfile if present, otherwise Google Cloud buildpacks (Go, Node.js, Python, Java, .NET, Ruby, PHP); deploy-without-build is Preview | |
| ||||||||||
| Yes. Per-app cron jobs with run history, logs, retries and alerts | Yes. Scheduled tasks run commands in containers on cron schedules | Yes. Schedule Jobs runs cron jobs in apps, compose services or servers | Not stated in their docs. | Yes. Scheduled dokku run tasks via cron:set and app.json | Not stated in their docs. | Yes. Cron Jobs on all plans (UTC, run Vercel Functions); docs now mark Cron Jobs as legacy, moving to Schedules with timezones and runtime creation | Partly. Heroku Scheduler is a free add-on running every 10 minutes, hourly or daily (daily in UTC); jobs bill as one-off dyno hours; Heroku advises a custom clock process for critical jobs because runs can be skipped or duplicated | Yes. Cron schedule per service; runs at least 5 minutes apart, UTC | Partly. Jobs can run once, on a schedule or in a workflow; delayed job execution is in Preview | |
No answer in their docs or not applicable: CapRover, Kamal. | ||||||||||
| Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys | Partly. Image scanning (Syft and Grype, v4.7.0) has a per-instance/team/server/app 'Block deploys' policy (Off, On, Strict) with a 'Deploy anyway' override; it is a vulnerability gate, not deploy freezes or change windows | Not stated in their docs. | Not stated in their docs. | Partly. Per-app deploy lock (apps:lock) and checks:disable/skip exist; no freeze windows or policy engine documented | Partly. kamal lock acquire can block deploys during a maintenance window and a non-zero exit from pre-deploy or other hooks aborts the command (hooks can be skipped with --skip_hooks); no scheduled freeze windows or policy engine | Partly. Deployment Checks (native lint and typecheck, GitHub Actions results, Marketplace checks) hold production promotion until they pass; no freeze windows stated; Force Promote bypasses | Partly. 'Wait for CI to pass before deploy' holds GitHub auto-deploys until commit statuses succeed; no freeze windows stated | Partly. Wait for CI holds deploys until GitHub Actions check suites finish; Enterprise guardrails restrict deploy sources; no freeze windows documented | Partly. Binary Authorization policy lets Cloud Run deploy only images signed by trusted authorities, with allowlist patterns; freeze windows not stated | |
No answer in their docs or not applicable: Dokploy, CapRover. | ||||||||||
| Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | No. Official GitHub Action and Woodpecker examples use a static SSH private key stored as a CI secret; OIDC is not mentioned | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | No. No OIDC or keyless CI deploy documented; deploys use project or account tokens stored as CI secrets | Yes. Workload Identity Federation lets GitHub Actions and GitLab CI jobs use their OIDC tokens for Google Cloud credentials, without service account keys | |
No answer in their docs or not applicable: Coolify, Dokploy, CapRover, Kamal, Vercel, Heroku. | ||||||||||
| Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy | Yes. Static build pack serves committed files through Nginx; Nixpacks or Railpack with Static site type and publish directory for source projects | Yes. Static build type serves files through NGINX; Nixpacks publish directory for built output | Partly. Documented guide: build locally, ship an NGINX Dockerfile and captain-definition; no dedicated static build mode | Not stated in their docs. | Not stated in their docs. | Yes. Framework auto-detected builds served from the CDN; static assets are counted as CDN requests and Hobby includes 1M of them | Not stated in their docs. | Yes. Zero-config static site deploys from GitHub with custom domains and automatic SSL | Not stated in their docs. | |
No answer in their docs or not applicable: Dokku, Kamal, Heroku, Cloud Run. | ||||||||||
| Data | ||||||||||
| Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta | Yes. Scheduled backups to S3-compatible storage for six database engines | Yes. Creates MySQL, Postgres, MongoDB, Redis, MariaDB; scheduled backups | Partly. One-click databases; docs tell you to schedule your own database-aware backups; built-in backup covers CapRover config only; no S3 mentioned | Partly. Official plugins create databases; backups manual export, no scheduling | Partly. Accessories run containers such as MySQL; no backups documented | Partly. Marketplace databases from the dashboard; backup schedules not stated | Partly. Postgres Essential-0 $5 (1 GB, 20 connections, no rollback, no fork/follow, 99.5% uptime); daily logical backups; Advanced tier (usage-based, from $150, HA by default) replaces Classic and Heroku PGBackups is not supported on Advanced | Partly. Scheduled volume and Postgres PITR backups; S3 export needs own job | Not stated in their docs. | |
No answer in their docs or not applicable: Cloud Run. | ||||||||||
| Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore | No. Scheduled engine-aware dumps with S3 copy; no continuous archiving or restore to a timestamp is described | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Partly. Rollback window by tier: Standard 4 days, Premium/Private/Shield/Advanced 7 days; Essential has no rollback | Not stated in their docs. | Not stated in their docs. | |
No answer in their docs or not applicable: Dokploy, CapRover, Dokku, Kamal, Vercel, Railway, Cloud Run. | ||||||||||
| Yes. Envelope-encrypted, masked in UI, injected at container create | Partly. Doppler, Infisical and HashiCorp Vault secret managers documented; encryption of ordinary env vars at rest not stated; v4.6.0 API/MCP can manage secret managers | Partly. External secrets providers fetch values at deploy time; encryption not stated | Not stated in their docs. | Not stated in their docs. | Partly. kamal secrets helpers extract secrets from password managers; custom TLS certs load from secrets; no encryption at rest documented | Yes. Encrypted at rest; Secret type values are write-only after saving | Not stated in their docs. | Yes. Variables encrypted at rest; sealed values write-only | Yes. Secret Manager values injected as env vars or mounted volumes | |
No answer in their docs or not applicable: CapRover, Dokku, Heroku. | ||||||||||
| Networking | ||||||||||
| Active. Named networks and allow rules; cross-server mesh unverified on real hosts | Partly. Shared Docker networks on one server; cross-server networking not documented | Not stated in their docs. | Yes. Swarm overlay network; internal traffic uses the app service name and container port | Partly. Docker networks on one host only; no cross-host networking | Partly. Containers share a custom Docker network per host, no cross server networking | Partly. Secure Compute private networks and VPC peering, Enterprise only | Partly. Private Spaces, Teams or Enterprise only: internal routing and DNS | Yes. WireGuard private DNS per environment, service.railway.internal | Partly. Direct VPC egress reaches private IPs; internal DNS not documented | |
No answer in their docs or not applicable: Dokploy. | ||||||||||
| Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only | Partly. Traffic balancing needs an external cloud load balancer, set up manually | Yes. unchanged | Partly. Swarm schedules and routes stateless replicas; the docs do not describe load balancing directly; persistent apps are limited to one instance | Partly. Nginx round-robin across scaled web containers; other processes need config | No. Proxy config covers hosts, health checks, timeouts and buffering for one server; no balancing across servers is described | Not applicable. | Yes. Router randomly balances requests across web dynos | Yes. Traffic spread across replicas and regions; no sticky sessions | Yes. Requests balanced across regional instances automatically; multi region needs load balancer | |
No answer in their docs or not applicable: Vercel. | ||||||||||
| Yes. Ordered rules, IP bans, bot handling, challenges and rate limits | Partly. Rate limits via custom Traefik middleware; no bot protection | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | No. Proxy options cover buffering, body size limit and timeouts; no rate limits or bot protection described | Partly. DDoS mitigation on all plans; WAF on all plans with 3 custom rules on Hobby, 40 on Pro, 250 on Enterprise; Bot Protection and managed rulesets; managed OWASP ruleset is Enterprise and custom priced | Not stated in their docs. | Partly. Under Attack Mode plus Edge rules (block, allow, challenge, redirect, cache override); rate limiting not documented | Partly. Cloud Armor filters traffic only through an external load balancer | |
No answer in their docs or not applicable: Dokploy, CapRover, Dokku, Heroku. | ||||||||||
| Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported | Yes. Traefik or Caddy requests and renews certificates for https domains | Yes. Per-domain HTTPS with a Let's Encrypt certificate resolver option | Yes. Let's Encrypt certificates managed automatically by CapRover | Partly. Let's Encrypt via official plugin; manual certs otherwise | Partly. kamal-proxy Let's Encrypt HTTPS, only with one server and host set | Yes. Let's Encrypt certs issued and renewed automatically per domain | Yes. Automated Certificate Management: free Let's Encrypt certs for custom domains | Yes. Let's Encrypt certificates for custom domains, renewed automatically | Partly. Domain mapping gives auto-issued and renewed Google-managed certs but is Preview and not recommended for production; global external Application Load Balancer is the recommended path. New Preview (Sep 28) custom URLs like example.cloud.run need no DNS or load balancer | |
| ||||||||||
| Platform | ||||||||||
| Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta | Yes. Manages each server over SSH | Yes. Multi-node scheduling through Docker Swarm | Yes. Add manager or worker nodes in Cluster > Nodes with SSH key and a user with Docker access | Partly. One host by default; optional k3s scheduler joins worker and server nodes over SSH with scheduler-k3s:cluster:add | Yes. SSH to each server | Not applicable. | Not applicable. | Not applicable. | Not applicable. | |
No answer in their docs or not applicable: Vercel, Heroku, Railway, Cloud Run. | ||||||||||
| Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet | Yes. Large catalog of one-click services | Yes. 602 or more one-click open source templates | Yes. One-click apps catalog in the dashboard; third-party template repositories supported; count not stated in docs | Not stated in their docs. | Not stated in their docs. | Partly. Framework starter templates listed; no service template catalog stated | Partly. app.json and Deploy button; no one-click service catalog | Yes. Template marketplace with one-click deploys | Not stated in their docs. | |
No answer in their docs or not applicable: Dokku, Kamal, Cloud Run. | ||||||||||
| Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons | No. Scaling is manual; operators decide when to add servers | Not stated in their docs. | No. Replica count is set by hand in the dashboard | Partly. docker-local scheduler scales manually via ps:scale; the optional k3s scheduler documents autoscaling through KEDA | Not stated in their docs. | Yes. Fluid compute scales instances automatically with traffic | Partly. Web dynos only, Performance, Private or Shield tiers, scaled on a p95 response time target; not available for Fir apps | No. Vertical scaling is automatic per container, but replica count stays where you set it; horizontal autoscaling means building your own with the API, railway scale CLI or cron | Yes. Scales on requests, events and CPU use; min and max instances; manual scaling available; custom CPU or concurrency utilization targets reached GA on 2026-09-29 | |
No answer in their docs or not applicable: Dokploy, Kamal. | ||||||||||
| Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Partly. Eco web dynos sleep after 30 minutes without web traffic; other dyno types do not | Yes. Serverless setting per service; sleeps 5 to 10 minutes after last outbound traffic, wakes on inbound request, first request may be slow or return 502 | Yes. Scales to zero instances by default when there is no traffic | |
No answer in their docs or not applicable: Coolify, Dokploy, CapRover, Dokku, Kamal, Vercel. | ||||||||||
| Yes. CLI and REST API documented; MCP server is beta | Partly. REST API with scoped tokens (read to root, delete ability since v4.6.0) and a CLI built on it; docs do not claim full dashboard parity | Partly. HTTP API and a CLI; CLI covers app create, deploy, stop, delete | Partly. CLI: serversetup, login, list, deploy, logout, api; HTTP API under /api/v2 with route groups for apps, projects, one-click, registries, system; per-app deploy tokens for CI | Partly. CLI over SSH; JSON API only in paid Dokku Pro | Not applicable. | Yes. CLI covers deploy, logs, env, promote; REST API also available | Partly. CLI and Platform API for apps and config; full parity not stated | Yes. Public GraphQL API is the same API that powers the dashboard; CLI and MCP server also available | Yes. gcloud CLI and Admin API manage services and jobs | |
No answer in their docs or not applicable: Kamal. | ||||||||||
| Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes | Yes. Email, Discord, Telegram, Slack, Pushover and webhook channels | Yes. Slack, Discord, Email, Telegram, Teams and Webhook notifications | Partly. Build status notifications are Pro; v1.16.0 adds persistent low-disk alerts; channels not named | Partly. Deploy webhook via community plugin; no built-in notifications | Partly. Hooks run custom scripts; no built-in Slack, Discord or email | Partly. Deployment failure notices by dashboard, email and push on all plans; anomaly alerts by email, Slack or webhook, but those need Pro or Enterprise with Observability Plus | Partly. Threshold alerts on p95 response time or 5xx rate by email, PagerDuty or dashboard; Slack through ChatOps; webhooks via Platform API | Yes. Slack and Discord webhooks; email via account notifications | Partly. Cloud Monitoring alerts to email, Slack, webhooks, Pub/Sub | |
| ||||||||||
| Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag | Yes. MCP server over Streamable HTTP with team-scoped tokens | Yes. Official Dokploy/mcp package: 508 tools generated from the OpenAPI spec, stdio and Streamable HTTP, secret redaction by default. Documented in the GitHub repo, not the docs site | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Yes. Official remote Vercel MCP server at mcp.vercel.com, OAuth | Yes. Remote MCP server at mcp.heroku.com with OAuth login | Yes. Official hosted MCP server at mcp.railway.com | Partly. Remote MCP server (Streamable HTTP, OAuth 2.0 and IAM) deploys from images, source or zip and lists and inspects services; regional endpoints and the /deploy prompt are Preview; Cloud Audit Logs and optional Model Armor | |
No answer in their docs or not applicable: CapRover, Dokku, Kamal. | ||||||||||
| Yes. Runs on Linux servers you provide | Yes. Web application on your server | Yes. Web application on your server | Yes. Web application running as a Swarm service | Yes. Scripts running on your own host | Yes. Runs against servers you own over SSH | Not applicable. | Not applicable. | Not applicable. | Not applicable. | |
No answer in their docs or not applicable: Vercel, Heroku, Railway, Cloud Run. | ||||||||||
| Yes. Apache 2.0, one edition, no paid tier | Yes. Listed as open source | Yes. Apache-2.0 core with a proprietary source-available directory; enterprise features need a license key | Yes. Apache-2.0 plus an appendix reserving paid Pro features | Yes. Permissive MIT-style license per repo LICENSE file | Yes. Released under the MIT License | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | |
No answer in their docs or not applicable: Vercel, Heroku, Railway, Cloud Run. | ||||||||||
| Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Partly. app.json declares per-app health checks, cron, formation and env; there is no whole-platform spec to apply | Partly. config/deploy.yml declares app, roles, proxy, accessories and servers and kamal setup/deploy applies it; commands run per deploy, no state reconciliation or drift detection documented | Partly. Per-project config in vercel.json or vercel.ts; WAF can be managed through REST API, SDK or Terraform; no whole-platform declarative apply stated | Partly. Terraform listed as a deploy path and app.json for app setup; Platform API manages apps, pipelines, config vars and add-ons; no whole-platform declarative apply stated | Yes. Infrastructure as Code in .railway/railway.ts (GA; Python and Go in beta) defines services, databases, volumes, buckets, domains, variables, replicas; railway config plan and apply; railway.json/toml Config as Code is deprecated, legacy files read until 2026-12-01 | Partly. Per service Knative YAML export and 'gcloud run services replace', plus Terraform google_cloud_run_v2_service; whole-platform apply is Terraform across Google Cloud | |
No answer in their docs or not applicable: Coolify, Dokploy, CapRover. | ||||||||||
| Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries | Partly. Scheduled tasks run commands in containers on cron with execution history; no functions runtime documented | Partly. Schedule Jobs run cron commands in apps, compose services, remote hosts or the Dokploy container with run logs; no functions runtime documented | Not stated in their docs. | Partly. Scheduled cron tasks in app.json run as one-off containers; a lambda builder exists for AWS Lambda functions; no functions runtime | Partly. A role can run a custom cmd such as bin/jobs for a job server; scheduled jobs are not described | Yes. Vercel Functions on Fluid compute, Cron Jobs on all plans, plus Queues, Workflow and Sandbox listed as background options | Partly. Worker dynos, one-off dynos and the free Scheduler add-on (10 minute, hourly or daily); no function-style runtime stated | Yes. Cron services plus Functions (single TypeScript file on Bun, with optional cron schedule) | Yes. Services plus Cloud Run jobs (up to 10,000 parallel tasks, schedule or workflow runs); sandboxes for jobs and worker pools in Preview | |
No answer in their docs or not applicable: CapRover. | ||||||||||
| Security | ||||||||||
| Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition | Yes. Owner/Admin/Member roles, team audit log (90 day retention) and 2FA documented; docs and pricing say self-hosted has the same features, no paid tier | Partly. Owner, Admin, Member roles and 2FA/passkeys in self-hosted; custom roles and audit logs need an Enterprise license | Partly. 2FA and login alerts are Pro; no roles or audit log documented | Partly. SSH key users; admin keyword grants key management; no RBAC | Partly. Per-server audit log (.kamal/app-audit.log) of deploy steps, last 50 lines via kamal audit; no users, roles or 2FA | Partly. Team roles on Pro and Enterprise; project-level roles, Security and Contributor roles and audit logs are Enterprise only; no audit logs on Hobby or Pro | Partly. Four team roles and mandatory MFA; audit trails Enterprise only | Partly. Audit logs: 48 hours on Free, Trial and Hobby, 30 days on Pro, 18 months on Enterprise; Environment RBAC and SAML SSO are Enterprise or committed-spend only | Partly. Cloud IAM and audit logs; 2FA is not a Cloud Run feature | |
| ||||||||||
| Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra | Partly. OpenID Connect sign-in documented; SAML and SCIM not mentioned | Yes. SAML and OIDC SSO plus SCIM, under the Enterprise license | Not stated in their docs. | Not stated in their docs. | Not applicable. | Partly. SAML SSO on Pro as a $300/month add-on and included on Enterprise; Directory Sync (SCIM) on Enterprise only | Partly. SAML SSO for Heroku Teams and Enterprise; accounts auto-provisioned on first IdP login, removal needs Support; SCIM not documented | Partly. SAML SSO is an Enterprise feature; SCIM not mentioned | Not stated in their docs. | |
No answer in their docs or not applicable: CapRover, Dokku, Kamal, Cloud Run. | ||||||||||
| Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket | Yes. Team audit log of resource changes, deployments, backups, membership, API tokens and settings from UI, API, MCP and webhooks; admins and owners; events pruned after 90 days; no plan restriction stated | Partly. Enterprise only; v0.31.0 adds export to CSV, TSV or JSON (newest 10,000 per export); retention not stated | Not stated in their docs. | Partly. Optional event logger (events:on) records plugin trigger invocations to syslog and /var/log/dokku/events.log; page does not show who ran a command or a retention period | Partly. Per-server .kamal/app-audit.log of deploy steps, read with kamal audit; entries carry a user tag but the page does not say which person; hooks expose KAMAL_PERFORMER for custom reports | Partly. Audit Logs are Enterprise only, owner and security roles; CSV export covers the last 90 days. Custom SIEM streaming is being replaced by Audit Log Drains | Partly. Audit trail export (JSON archive per month) for Enterprise accounts only, needs manage permission | Yes. Retention 48 hours on Free, Trial and Hobby, 30 days on Pro, 18 months on Enterprise; viewable by workspace admins | Partly. Cloud Audit Logs cover Cloud Run API methods; the page does not state which are on by default, and Data Access logs appear to need enabling; pricing and retention not stated on the page | |
No answer in their docs or not applicable: CapRover. | ||||||||||
| Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Not stated in their docs. | Partly. Deployment Approvals hold commits from users without a linked Railway account; duplicated environments need staged-change review; Environment RBAC and approved-sources guardrail are Enterprise or committed-spend; no two-person production approval documented | Partly. Via Cloud Deploy, a separate product: requireApproval on a target, approver role roles/clouddeploy.approver. The page opened describes Cloud Deploy generally, not a Cloud Run page | |
No answer in their docs or not applicable: Coolify, Dokploy, CapRover, Dokku, Kamal, Vercel, Heroku. | ||||||||||
Pick a platform to see it next to Levelrail, row by row.
Architecture
No SSH: agents dial out
Levelrail Yes. Agents dial out over mTLS; no inbound ports Source for Levelrail, No SSH: agents dial out
Coolify No. SSH from the control plane to each server Source for Coolify, No SSH: agents dial out
Level-triggered reconciler with status reasons
Levelrail Yes. Desired state converged, status condition with reason per pass Source for Levelrail, Level-triggered reconciler with status reasons
Coolify Not stated in their docs
Single control plane binary
Levelrail Yes. One control plane binary, plus one agent per extra node Source for Levelrail, Single control plane binary
Coolify No. Three base containers: app, PostgreSQL and Redis Source for Coolify, Single control plane binary
Observability
Metrics and log search built in
Levelrail Yes. Node-local metrics at 15s and full-text log search Source for Levelrail, Metrics and log search built in
Coolify Partly. Sentinel historical CPU, memory, disk, network and load charts; log drains to Axiom, New Relic, CloudWatch (new in v4.6.0), Fluent Bit; docs say log queries and dashboards need an external system Source for Coolify, Metrics and log search built in
Crashloop diagnosis with last log lines
Levelrail Yes. Crashloop detection surfaces the last log lines Source for Levelrail, Crashloop diagnosis with last log lines
Coolify Not stated in their docs
Public status page
Levelrail Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS Source for Levelrail, Public status page
Coolify Not stated in their docs
Deploys
One-click rollback with pinned images
Levelrail Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta Source for Levelrail, One-click rollback with pinned images
Coolify Yes. Redeploys a retained older image; retention is configurable Source for Coolify, One-click rollback with pinned images
Readiness-gated deploys (zero-downtime on one node)
Levelrail Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta Source for Levelrail, Readiness-gated deploys (zero-downtime on one node)
Coolify Partly. Starts replacement first; docs say no zero-downtime guarantee Source for Coolify, Readiness-gated deploys (zero-downtime on one node)
Preview environments per pull request
Levelrail Yes. Stable on GitHub, beta on GitLab and Bitbucket Source for Levelrail, Preview environments per pull request
Coolify Yes. Per pull request previews; full UI needs a GitHub App Source for Coolify, Preview environments per pull request
Docker Compose apps
Levelrail Active. Pre-built images from the dashboard; git deploys can build Source for Levelrail, Docker Compose apps
Coolify Yes. Deploys docker-compose.yml from a git repository or pasted file Source for Coolify, Docker Compose apps
Git push deploys from GitHub, GitLab, Bitbucket
Levelrail Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta Source for Levelrail, Git push deploys from GitHub, GitLab, Bitbucket
Coolify Yes. Push webhooks for GitHub, GitLab and Bitbucket sources Source for Coolify, Git push deploys from GitHub, GitLab, Bitbucket
Builds from source without a Dockerfile
Levelrail Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected Source for Levelrail, Builds from source without a Dockerfile
Coolify Yes. Nixpacks and Railpack detect and build without a Dockerfile Source for Coolify, Builds from source without a Dockerfile
Scheduled jobs
Levelrail Yes. Per-app cron jobs with run history, logs, retries and alerts Source for Levelrail, Scheduled jobs
Coolify Yes. Scheduled tasks run commands in containers on cron schedules Source for Coolify, Scheduled jobs
Deploy freezes and policies
Levelrail Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys Source for Levelrail, Deploy freezes and policies
Coolify Partly. Image scanning (Syft and Grype, v4.7.0) has a per-instance/team/server/app 'Block deploys' policy (Off, On, Strict) with a 'Deploy anyway' override; it is a vulnerability gate, not deploy freezes or change windows Source for Coolify, Deploy freezes and policies
CI deploys with no stored secret
Levelrail Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token Source for Levelrail, CI deploys with no stored secret
Coolify Not stated in their docs
Static site hosting
Levelrail Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy Source for Levelrail, Static site hosting
Coolify Yes. Static build pack serves committed files through Nginx; Nixpacks or Railpack with Static site type and publish directory for source projects Source for Coolify, Static site hosting
Data
Managed databases with scheduled backups
Levelrail Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta Source for Levelrail, Managed databases with scheduled backups
Coolify Yes. Scheduled backups to S3-compatible storage for six database engines Source for Coolify, Managed databases with scheduled backups
Point-in-time database restore
Levelrail Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore Source for Levelrail, Point-in-time database restore
Coolify No. Scheduled engine-aware dumps with S3 copy; no continuous archiving or restore to a timestamp is described Source for Coolify, Point-in-time database restore
Encrypted secrets for env vars
Levelrail Yes. Envelope-encrypted, masked in UI, injected at container create Source for Levelrail, Encrypted secrets for env vars
Coolify Partly. Doppler, Infisical and HashiCorp Vault secret managers documented; encryption of ordinary env vars at rest not stated; v4.6.0 API/MCP can manage secret managers Source for Coolify, Encrypted secrets for env vars
Networking
Private networking between apps or nodes
Levelrail Active. Named networks and allow rules; cross-server mesh unverified on real hosts Source for Levelrail, Private networking between apps or nodes
Coolify Partly. Shared Docker networks on one server; cross-server networking not documented Source for Coolify, Private networking between apps or nodes
Load balancing across replicas
Levelrail Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only Source for Levelrail, Load balancing across replicas
Coolify Partly. Traffic balancing needs an external cloud load balancer, set up manually Source for Coolify, Load balancing across replicas
Request firewall, rate limits, bot protection
Levelrail Yes. Ordered rules, IP bans, bot handling, challenges and rate limits Source for Levelrail, Request firewall, rate limits, bot protection
Coolify Partly. Rate limits via custom Traefik middleware; no bot protection Source for Coolify, Request firewall, rate limits, bot protection
Automatic TLS certificates
Levelrail Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported Source for Levelrail, Automatic TLS certificates
Coolify Yes. Traefik or Caddy requests and renews certificates for https domains Source for Coolify, Automatic TLS certificates
Platform
Multi-node
Levelrail Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta Source for Levelrail, Multi-node
Coolify Yes. Manages each server over SSH Source for Coolify, Multi-node
One-click service templates
Levelrail Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet Source for Levelrail, One-click service templates
Coolify Yes. Large catalog of one-click services Source for Coolify, One-click service templates
Autoscaling
Levelrail Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons Source for Levelrail, Autoscaling
Coolify No. Scaling is manual; operators decide when to add servers Source for Coolify, Autoscaling
Sleep idle apps, wake on request
Levelrail Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress Source for Levelrail, Sleep idle apps, wake on request
Coolify Not stated in their docs
CLI and HTTP API
Levelrail Yes. CLI and REST API documented; MCP server is beta Source for Levelrail, CLI and HTTP API
Coolify Partly. REST API with scoped tokens (read to root, delete ability since v4.6.0) and a CLI built on it; docs do not claim full dashboard parity Source for Coolify, CLI and HTTP API
Deploy and alert notifications
Levelrail Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes Source for Levelrail, Deploy and alert notifications
Coolify Yes. Email, Discord, Telegram, Slack, Pushover and webhook channels Source for Coolify, Deploy and alert notifications
MCP / AI assistant integration
Levelrail Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag Source for Levelrail, MCP / AI assistant integration
Coolify Yes. MCP server over Streamable HTTP with team-scoped tokens Source for Coolify, MCP / AI assistant integration
Runs on servers you own
Levelrail Yes. Runs on Linux servers you provide Source for Levelrail, Runs on servers you own
Coolify Yes. Web application on your server Source for Coolify, Runs on servers you own
Open source license
Levelrail Yes. Apache 2.0, one edition, no paid tier Source for Levelrail, Open source license
Coolify Yes. Listed as open source Source for Coolify, Open source license
Declarative apply of the whole platform
Levelrail Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag Source for Levelrail, Declarative apply of the whole platform
Coolify Not stated in their docs
Functions and one-off jobs
Levelrail Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries Source for Levelrail, Functions and one-off jobs
Coolify Partly. Scheduled tasks run commands in containers on cron with execution history; no functions runtime documented Source for Coolify, Functions and one-off jobs
Security
IAM, audit log and 2FA in the free edition
Levelrail Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition Source for Levelrail, IAM, audit log and 2FA in the free edition
Coolify Yes. Owner/Admin/Member roles, team audit log (90 day retention) and 2FA documented; docs and pricing say self-hosted has the same features, no paid tier Source for Coolify, IAM, audit log and 2FA in the free edition
SAML or SCIM single sign-on
Levelrail Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra Source for Levelrail, SAML or SCIM single sign-on
Coolify Partly. OpenID Connect sign-in documented; SAML and SCIM not mentioned Source for Coolify, SAML or SCIM single sign-on
Audit log of every action
Levelrail Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket Source for Levelrail, Audit log of every action
Coolify Yes. Team audit log of resource changes, deployments, backups, membership, API tokens and settings from UI, API, MCP and webhooks; admins and owners; events pruned after 90 days; no plan restriction stated Source for Coolify, Audit log of every action
Two-person deploy approvals
Levelrail Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord Source for Levelrail, Two-person deploy approvals
Coolify Not stated in their docs
Architecture
No SSH: agents dial out
Levelrail Yes. Agents dial out over mTLS; no inbound ports Source for Levelrail, No SSH: agents dial out
Dokploy No. Connects to remote servers over SSH keys Source for Dokploy, No SSH: agents dial out
Level-triggered reconciler with status reasons
Levelrail Yes. Desired state converged, status condition with reason per pass Source for Levelrail, Level-triggered reconciler with status reasons
Dokploy Not stated in their docs
Single control plane binary
Levelrail Yes. One control plane binary, plus one agent per extra node Source for Levelrail, Single control plane binary
Dokploy No. Three parts: Next.js app, PostgreSQL and Traefik Source for Dokploy, Single control plane binary
Observability
Metrics and log search built in
Levelrail Yes. Node-local metrics at 15s and full-text log search Source for Levelrail, Metrics and log search built in
Dokploy Partly. CPU, memory, disk, network metrics with threshold alerts, real-time logs, and v0.31.0 Vector shipping of logs and metrics to external providers; log search not stated Source for Dokploy, Metrics and log search built in
Crashloop diagnosis with last log lines
Levelrail Yes. Crashloop detection surfaces the last log lines Source for Levelrail, Crashloop diagnosis with last log lines
Dokploy Not stated in their docs
Public status page
Levelrail Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS Source for Levelrail, Public status page
Dokploy Not stated in their docs
Deploys
One-click rollback with pinned images
Levelrail Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta Source for Levelrail, One-click rollback with pinned images
Dokploy Yes. Registry-based rollback to any previous deployment image (needs a configured registry); Swarm automatic rollback on failed health check Source for Dokploy, One-click rollback with pinned images
Readiness-gated deploys (zero-downtime on one node)
Levelrail Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta Source for Levelrail, Readiness-gated deploys (zero-downtime on one node)
Dokploy Partly. Swarm health check plus update config with automatic rollback on failure is documented (needs a /health endpoint and curl in the image) Source for Dokploy, Readiness-gated deploys (zero-downtime on one node)
Preview environments per pull request
Levelrail Yes. Stable on GitHub, beta on GitLab and Bitbucket Source for Levelrail, Preview environments per pull request
Dokploy Partly. Preview per pull request, documented for GitHub only Source for Dokploy, Preview environments per pull request
Docker Compose apps
Levelrail Active. Pre-built images from the dashboard; git deploys can build Source for Levelrail, Docker Compose apps
Dokploy Yes. Compose deploys from Git or pasted file; build unsupported in Stack mode Source for Dokploy, Docker Compose apps
Git push deploys from GitHub, GitLab, Bitbucket
Levelrail Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta Source for Levelrail, Git push deploys from GitHub, GitLab, Bitbucket
Dokploy Yes. Auto deploy on push for GitHub, GitLab, Bitbucket and Gitea Source for Dokploy, Git push deploys from GitHub, GitLab, Bitbucket
Builds from source without a Dockerfile
Levelrail Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected Source for Levelrail, Builds from source without a Dockerfile
Dokploy Yes. Nixpacks is the default; Railpack and buildpacks also offered Source for Dokploy, Builds from source without a Dockerfile
Scheduled jobs
Levelrail Yes. Per-app cron jobs with run history, logs, retries and alerts Source for Levelrail, Scheduled jobs
Dokploy Yes. Schedule Jobs runs cron jobs in apps, compose services or servers Source for Dokploy, Scheduled jobs
Deploy freezes and policies
Levelrail Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys Source for Levelrail, Deploy freezes and policies
Dokploy Not stated in their docs
CI deploys with no stored secret
Levelrail Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token Source for Levelrail, CI deploys with no stored secret
Dokploy Not stated in their docs
Static site hosting
Levelrail Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy Source for Levelrail, Static site hosting
Dokploy Yes. Static build type serves files through NGINX; Nixpacks publish directory for built output Source for Dokploy, Static site hosting
Data
Managed databases with scheduled backups
Levelrail Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta Source for Levelrail, Managed databases with scheduled backups
Dokploy Yes. Creates MySQL, Postgres, MongoDB, Redis, MariaDB; scheduled backups Source for Dokploy, Managed databases with scheduled backups
Point-in-time database restore
Levelrail Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore Source for Levelrail, Point-in-time database restore
Dokploy Not stated in their docs
Encrypted secrets for env vars
Levelrail Yes. Envelope-encrypted, masked in UI, injected at container create Source for Levelrail, Encrypted secrets for env vars
Dokploy Partly. External secrets providers fetch values at deploy time; encryption not stated Source for Dokploy, Encrypted secrets for env vars
Networking
Private networking between apps or nodes
Levelrail Active. Named networks and allow rules; cross-server mesh unverified on real hosts Source for Levelrail, Private networking between apps or nodes
Dokploy Not stated in their docs
Load balancing across replicas
Levelrail Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only Source for Levelrail, Load balancing across replicas
Dokploy Yes. unchanged Source for Dokploy, Load balancing across replicas
Request firewall, rate limits, bot protection
Levelrail Yes. Ordered rules, IP bans, bot handling, challenges and rate limits Source for Levelrail, Request firewall, rate limits, bot protection
Dokploy Not stated in their docs
Automatic TLS certificates
Levelrail Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported Source for Levelrail, Automatic TLS certificates
Dokploy Yes. Per-domain HTTPS with a Let's Encrypt certificate resolver option Source for Dokploy, Automatic TLS certificates
Platform
Multi-node
Levelrail Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta Source for Levelrail, Multi-node
Dokploy Yes. Multi-node scheduling through Docker Swarm Source for Dokploy, Multi-node
One-click service templates
Levelrail Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet Source for Levelrail, One-click service templates
Dokploy Yes. 602 or more one-click open source templates Source for Dokploy, One-click service templates
Autoscaling
Levelrail Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons Source for Levelrail, Autoscaling
Dokploy Not stated in their docs
Sleep idle apps, wake on request
Levelrail Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress Source for Levelrail, Sleep idle apps, wake on request
Dokploy Not stated in their docs
CLI and HTTP API
Levelrail Yes. CLI and REST API documented; MCP server is beta Source for Levelrail, CLI and HTTP API
Dokploy Partly. HTTP API and a CLI; CLI covers app create, deploy, stop, delete Source for Dokploy, CLI and HTTP API
Deploy and alert notifications
Levelrail Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes Source for Levelrail, Deploy and alert notifications
Dokploy Yes. Slack, Discord, Email, Telegram, Teams and Webhook notifications Source for Dokploy, Deploy and alert notifications
MCP / AI assistant integration
Levelrail Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag Source for Levelrail, MCP / AI assistant integration
Dokploy Yes. Official Dokploy/mcp package: 508 tools generated from the OpenAPI spec, stdio and Streamable HTTP, secret redaction by default. Documented in the GitHub repo, not the docs site Source for Dokploy, MCP / AI assistant integration
Runs on servers you own
Levelrail Yes. Runs on Linux servers you provide Source for Levelrail, Runs on servers you own
Dokploy Yes. Web application on your server Source for Dokploy, Runs on servers you own
Open source license
Levelrail Yes. Apache 2.0, one edition, no paid tier Source for Levelrail, Open source license
Dokploy Yes. Apache-2.0 core with a proprietary source-available directory; enterprise features need a license key Source for Dokploy, Open source license
Declarative apply of the whole platform
Levelrail Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag Source for Levelrail, Declarative apply of the whole platform
Dokploy Not stated in their docs
Functions and one-off jobs
Levelrail Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries Source for Levelrail, Functions and one-off jobs
Dokploy Partly. Schedule Jobs run cron commands in apps, compose services, remote hosts or the Dokploy container with run logs; no functions runtime documented Source for Dokploy, Functions and one-off jobs
Security
IAM, audit log and 2FA in the free edition
Levelrail Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition Source for Levelrail, IAM, audit log and 2FA in the free edition
Dokploy Partly. Owner, Admin, Member roles and 2FA/passkeys in self-hosted; custom roles and audit logs need an Enterprise license Source for Dokploy, IAM, audit log and 2FA in the free edition
SAML or SCIM single sign-on
Levelrail Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra Source for Levelrail, SAML or SCIM single sign-on
Dokploy Yes. SAML and OIDC SSO plus SCIM, under the Enterprise license Source for Dokploy, SAML or SCIM single sign-on
Audit log of every action
Levelrail Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket Source for Levelrail, Audit log of every action
Dokploy Partly. Enterprise only; v0.31.0 adds export to CSV, TSV or JSON (newest 10,000 per export); retention not stated Source for Dokploy, Audit log of every action
Two-person deploy approvals
Levelrail Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord Source for Levelrail, Two-person deploy approvals
Dokploy Not stated in their docs
Architecture
No SSH: agents dial out
Levelrail Yes. Agents dial out over mTLS; no inbound ports Source for Levelrail, No SSH: agents dial out
CapRover No. Nodes added with an SSH key from the dashboard Source for CapRover, No SSH: agents dial out
Level-triggered reconciler with status reasons
Levelrail Yes. Desired state converged, status condition with reason per pass Source for Levelrail, Level-triggered reconciler with status reasons
CapRover Not stated in their docs
Single control plane binary
Levelrail Yes. One control plane binary, plus one agent per extra node Source for Levelrail, Single control plane binary
CapRover Not stated in their docs
Observability
Metrics and log search built in
Levelrail Yes. Node-local metrics at 15s and full-text log search Source for Levelrail, Metrics and log search built in
CapRover Partly. Optional NetData for CPU, memory, disk, network; app and build logs viewable, no search documented; v1.16.0 adds runtime health, task history and disk alerts Source for CapRover, Metrics and log search built in
Crashloop diagnosis with last log lines
Levelrail Yes. Crashloop detection surfaces the last log lines Source for Levelrail, Crashloop diagnosis with last log lines
CapRover Not stated in their docs
Public status page
Levelrail Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS Source for Levelrail, Public status page
CapRover Not stated in their docs
Deploys
One-click rollback with pinned images
Levelrail Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta Source for Levelrail, One-click rollback with pinned images
CapRover Yes. Version history (up to 50 saved versions); image rollback does not restore env var values Source for CapRover, One-click rollback with pinned images
Readiness-gated deploys (zero-downtime on one node)
Levelrail Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta Source for Levelrail, Readiness-gated deploys (zero-downtime on one node)
CapRover Partly. Stateless apps default to start-first with optional health check; apps with mounted data default to stop-first; docs do not promise interruption-free updates Source for CapRover, Readiness-gated deploys (zero-downtime on one node)
Preview environments per pull request
Levelrail Yes. Stable on GitHub, beta on GitLab and Bitbucket Source for Levelrail, Preview environments per pull request
CapRover Not stated in their docs
Docker Compose apps
Levelrail Active. Pre-built images from the dashboard; git deploys can build Source for Levelrail, Docker Compose apps
CapRover Partly. Experimental import of a Compose subset (image, environment, ports, volumes, depends_on, command); build, networks, secrets, deploy ignored Source for CapRover, Docker Compose apps
Git push deploys from GitHub, GitLab, Bitbucket
Levelrail Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta Source for Levelrail, Git push deploys from GitHub, GitLab, Bitbucket
CapRover Partly. Generic push webhook documented with GitHub steps; other providers via their own webhook screens or CI Source for CapRover, Git push deploys from GitHub, GitLab, Bitbucket
Builds from source without a Dockerfile
Levelrail Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected Source for Levelrail, Builds from source without a Dockerfile
CapRover Partly. Dockerfile, prebuilt image, inline Dockerfile or built-in language templateId Source for CapRover, Builds from source without a Dockerfile
Scheduled jobs
Levelrail Yes. Per-app cron jobs with run history, logs, retries and alerts Source for Levelrail, Scheduled jobs
CapRover Not stated in their docs
Deploy freezes and policies
Levelrail Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys Source for Levelrail, Deploy freezes and policies
CapRover Not stated in their docs
CI deploys with no stored secret
Levelrail Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token Source for Levelrail, CI deploys with no stored secret
CapRover Not stated in their docs
Static site hosting
Levelrail Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy Source for Levelrail, Static site hosting
CapRover Partly. Documented guide: build locally, ship an NGINX Dockerfile and captain-definition; no dedicated static build mode Source for CapRover, Static site hosting
Data
Managed databases with scheduled backups
Levelrail Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta Source for Levelrail, Managed databases with scheduled backups
CapRover Partly. One-click databases; docs tell you to schedule your own database-aware backups; built-in backup covers CapRover config only; no S3 mentioned Source for CapRover, Managed databases with scheduled backups
Point-in-time database restore
Levelrail Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore Source for Levelrail, Point-in-time database restore
CapRover Not stated in their docs
Encrypted secrets for env vars
Levelrail Yes. Envelope-encrypted, masked in UI, injected at container create Source for Levelrail, Encrypted secrets for env vars
CapRover Not stated in their docs
Networking
Private networking between apps or nodes
Levelrail Active. Named networks and allow rules; cross-server mesh unverified on real hosts Source for Levelrail, Private networking between apps or nodes
CapRover Yes. Swarm overlay network; internal traffic uses the app service name and container port Source for CapRover, Private networking between apps or nodes
Load balancing across replicas
Levelrail Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only Source for Levelrail, Load balancing across replicas
CapRover Partly. Swarm schedules and routes stateless replicas; the docs do not describe load balancing directly; persistent apps are limited to one instance Source for CapRover, Load balancing across replicas
Request firewall, rate limits, bot protection
Levelrail Yes. Ordered rules, IP bans, bot handling, challenges and rate limits Source for Levelrail, Request firewall, rate limits, bot protection
CapRover Not stated in their docs
Automatic TLS certificates
Levelrail Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported Source for Levelrail, Automatic TLS certificates
CapRover Yes. Let's Encrypt certificates managed automatically by CapRover Source for CapRover, Automatic TLS certificates
Platform
Multi-node
Levelrail Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta Source for Levelrail, Multi-node
CapRover Yes. Add manager or worker nodes in Cluster > Nodes with SSH key and a user with Docker access Source for CapRover, Multi-node
One-click service templates
Levelrail Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet Source for Levelrail, One-click service templates
CapRover Yes. One-click apps catalog in the dashboard; third-party template repositories supported; count not stated in docs Source for CapRover, One-click service templates
Autoscaling
Levelrail Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons Source for Levelrail, Autoscaling
CapRover No. Replica count is set by hand in the dashboard Source for CapRover, Autoscaling
Sleep idle apps, wake on request
Levelrail Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress Source for Levelrail, Sleep idle apps, wake on request
CapRover Not stated in their docs
CLI and HTTP API
Levelrail Yes. CLI and REST API documented; MCP server is beta Source for Levelrail, CLI and HTTP API
CapRover Partly. CLI: serversetup, login, list, deploy, logout, api; HTTP API under /api/v2 with route groups for apps, projects, one-click, registries, system; per-app deploy tokens for CI Source for CapRover, CLI and HTTP API
Deploy and alert notifications
Levelrail Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes Source for Levelrail, Deploy and alert notifications
CapRover Partly. Build status notifications are Pro; v1.16.0 adds persistent low-disk alerts; channels not named Source for CapRover, Deploy and alert notifications
MCP / AI assistant integration
Levelrail Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag Source for Levelrail, MCP / AI assistant integration
CapRover Not stated in their docs
Runs on servers you own
Levelrail Yes. Runs on Linux servers you provide Source for Levelrail, Runs on servers you own
CapRover Yes. Web application running as a Swarm service Source for CapRover, Runs on servers you own
Open source license
Levelrail Yes. Apache 2.0, one edition, no paid tier Source for Levelrail, Open source license
CapRover Yes. Apache-2.0 plus an appendix reserving paid Pro features Source for CapRover, Open source license
Declarative apply of the whole platform
Levelrail Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag Source for Levelrail, Declarative apply of the whole platform
CapRover Not stated in their docs
Functions and one-off jobs
Levelrail Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries Source for Levelrail, Functions and one-off jobs
CapRover Not stated in their docs
Security
IAM, audit log and 2FA in the free edition
Levelrail Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition Source for Levelrail, IAM, audit log and 2FA in the free edition
CapRover Partly. 2FA and login alerts are Pro; no roles or audit log documented Source for CapRover, IAM, audit log and 2FA in the free edition
SAML or SCIM single sign-on
Levelrail Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra Source for Levelrail, SAML or SCIM single sign-on
CapRover Not stated in their docs
Audit log of every action
Levelrail Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket Source for Levelrail, Audit log of every action
CapRover Not stated in their docs
Two-person deploy approvals
Levelrail Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord Source for Levelrail, Two-person deploy approvals
CapRover Not stated in their docs
Architecture
No SSH: agents dial out
Levelrail Yes. Agents dial out over mTLS; no inbound ports Source for Levelrail, No SSH: agents dial out
Dokku No. Commands run on the host, over SSH or git push Source for Dokku, No SSH: agents dial out
Level-triggered reconciler with status reasons
Levelrail Yes. Desired state converged, status condition with reason per pass Source for Levelrail, Level-triggered reconciler with status reasons
Dokku No. Event-driven trigger pipeline; no reconcile loop described Source for Dokku, Level-triggered reconciler with status reasons
Single control plane binary
Levelrail Yes. One control plane binary, plus one agent per extra node Source for Levelrail, Single control plane binary
Dokku Not applicable
Observability
Metrics and log search built in
Levelrail Yes. Node-local metrics at 15s and full-text log search Source for Levelrail, Metrics and log search built in
Dokku No. Add your own stack Source for Dokku, Metrics and log search built in
Crashloop diagnosis with last log lines
Levelrail Yes. Crashloop detection surfaces the last log lines Source for Levelrail, Crashloop diagnosis with last log lines
Dokku Partly. Restart policy restarts failing containers; no crashloop detection Source for Dokku, Crashloop diagnosis with last log lines
Public status page
Levelrail Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS Source for Levelrail, Public status page
Dokku Not stated in their docs
Deploys
One-click rollback with pinned images
Levelrail Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta Source for Levelrail, One-click rollback with pinned images
Dokku Partly. Failed image deploys restore the previous image reference; old images are cleaned up unless skip-cleanup is set, and tags:deploy can redeploy a tagged image manually; no one-command rollback Source for Dokku, One-click rollback with pinned images
Readiness-gated deploys (zero-downtime on one node)
Levelrail Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta Source for Levelrail, Readiness-gated deploys (zero-downtime on one node)
Dokku Yes. New containers start, proxy switches, old retired after delay Source for Dokku, Readiness-gated deploys (zero-downtime on one node)
Preview environments per pull request
Levelrail Yes. Stable on GitHub, beta on GitLab and Bitbucket Source for Levelrail, Preview environments per pull request
Dokku Not stated in their docs
Docker Compose apps
Levelrail Active. Pre-built images from the dashboard; git deploys can build Source for Levelrail, Docker Compose apps
Dokku No. Deployment methods listed are git push, Dockerfile, Docker image and buildpacks; Compose files are not a documented deploy input Source for Dokku, Docker Compose apps
Git push deploys from GitHub, GitLab, Bitbucket
Levelrail Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta Source for Levelrail, Git push deploys from GitHub, GitLab, Bitbucket
Dokku Partly. git push to deploy branch; git:sync from any cloneable repo; GitHub push webhook deploys only in paid Dokku Pro 1.4.0; no GitLab or Bitbucket webhooks documented Source for Dokku, Git push deploys from GitHub, GitLab, Bitbucket
Builds from source without a Dockerfile
Levelrail Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected Source for Levelrail, Builds from source without a Dockerfile
Dokku Partly. Heroku buildpacks by default; pack and Railpack need config files Source for Dokku, Builds from source without a Dockerfile
Scheduled jobs
Levelrail Yes. Per-app cron jobs with run history, logs, retries and alerts Source for Levelrail, Scheduled jobs
Dokku Yes. Scheduled dokku run tasks via cron:set and app.json Source for Dokku, Scheduled jobs
Deploy freezes and policies
Levelrail Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys Source for Levelrail, Deploy freezes and policies
Dokku Partly. Per-app deploy lock (apps:lock) and checks:disable/skip exist; no freeze windows or policy engine documented Source for Dokku, Deploy freezes and policies
CI deploys with no stored secret
Levelrail Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token Source for Levelrail, CI deploys with no stored secret
Dokku No. Official GitHub Action and Woodpecker examples use a static SSH private key stored as a CI secret; OIDC is not mentioned Source for Dokku, CI deploys with no stored secret
Static site hosting
Levelrail Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy Source for Levelrail, Static site hosting
Dokku Not stated in their docs
Data
Managed databases with scheduled backups
Levelrail Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta Source for Levelrail, Managed databases with scheduled backups
Dokku Partly. Official plugins create databases; backups manual export, no scheduling Source for Dokku, Managed databases with scheduled backups
Point-in-time database restore
Levelrail Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore Source for Levelrail, Point-in-time database restore
Dokku Not stated in their docs
Encrypted secrets for env vars
Levelrail Yes. Envelope-encrypted, masked in UI, injected at container create Source for Levelrail, Encrypted secrets for env vars
Dokku Not stated in their docs
Networking
Private networking between apps or nodes
Levelrail Active. Named networks and allow rules; cross-server mesh unverified on real hosts Source for Levelrail, Private networking between apps or nodes
Dokku Partly. Docker networks on one host only; no cross-host networking Source for Dokku, Private networking between apps or nodes
Load balancing across replicas
Levelrail Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only Source for Levelrail, Load balancing across replicas
Dokku Partly. Nginx round-robin across scaled web containers; other processes need config Source for Dokku, Load balancing across replicas
Request firewall, rate limits, bot protection
Levelrail Yes. Ordered rules, IP bans, bot handling, challenges and rate limits Source for Levelrail, Request firewall, rate limits, bot protection
Dokku Not stated in their docs
Automatic TLS certificates
Levelrail Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported Source for Levelrail, Automatic TLS certificates
Dokku Partly. Let's Encrypt via official plugin; manual certs otherwise Source for Dokku, Automatic TLS certificates
Platform
Multi-node
Levelrail Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta Source for Levelrail, Multi-node
Dokku Partly. One host by default; optional k3s scheduler joins worker and server nodes over SSH with scheduler-k3s:cluster:add Source for Dokku, Multi-node
One-click service templates
Levelrail Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet Source for Levelrail, One-click service templates
Dokku Not stated in their docs
Autoscaling
Levelrail Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons Source for Levelrail, Autoscaling
Dokku Partly. docker-local scheduler scales manually via ps:scale; the optional k3s scheduler documents autoscaling through KEDA Source for Dokku, Autoscaling
Sleep idle apps, wake on request
Levelrail Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress Source for Levelrail, Sleep idle apps, wake on request
Dokku Not stated in their docs
CLI and HTTP API
Levelrail Yes. CLI and REST API documented; MCP server is beta Source for Levelrail, CLI and HTTP API
Dokku Partly. CLI over SSH; JSON API only in paid Dokku Pro Source for Dokku, CLI and HTTP API
Deploy and alert notifications
Levelrail Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes Source for Levelrail, Deploy and alert notifications
Dokku Partly. Deploy webhook via community plugin; no built-in notifications Source for Dokku, Deploy and alert notifications
MCP / AI assistant integration
Levelrail Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag Source for Levelrail, MCP / AI assistant integration
Dokku Not stated in their docs
Runs on servers you own
Levelrail Yes. Runs on Linux servers you provide Source for Levelrail, Runs on servers you own
Dokku Yes. Scripts running on your own host Source for Dokku, Runs on servers you own
Open source license
Levelrail Yes. Apache 2.0, one edition, no paid tier Source for Levelrail, Open source license
Dokku Yes. Permissive MIT-style license per repo LICENSE file Source for Dokku, Open source license
Declarative apply of the whole platform
Levelrail Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag Source for Levelrail, Declarative apply of the whole platform
Dokku Partly. app.json declares per-app health checks, cron, formation and env; there is no whole-platform spec to apply Source for Dokku, Declarative apply of the whole platform
Functions and one-off jobs
Levelrail Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries Source for Levelrail, Functions and one-off jobs
Dokku Partly. Scheduled cron tasks in app.json run as one-off containers; a lambda builder exists for AWS Lambda functions; no functions runtime Source for Dokku, Functions and one-off jobs
Security
IAM, audit log and 2FA in the free edition
Levelrail Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition Source for Levelrail, IAM, audit log and 2FA in the free edition
Dokku Partly. SSH key users; admin keyword grants key management; no RBAC Source for Dokku, IAM, audit log and 2FA in the free edition
SAML or SCIM single sign-on
Levelrail Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra Source for Levelrail, SAML or SCIM single sign-on
Dokku Not stated in their docs
Audit log of every action
Levelrail Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket Source for Levelrail, Audit log of every action
Dokku Partly. Optional event logger (events:on) records plugin trigger invocations to syslog and /var/log/dokku/events.log; page does not show who ran a command or a retention period Source for Dokku, Audit log of every action
Two-person deploy approvals
Levelrail Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord Source for Levelrail, Two-person deploy approvals
Dokku Not stated in their docs
Architecture
No SSH: agents dial out
Levelrail Yes. Agents dial out over mTLS; no inbound ports Source for Levelrail, No SSH: agents dial out
Kamal No. SSH from your machine or CI to each server Source for Kamal, No SSH: agents dial out
Level-triggered reconciler with status reasons
Levelrail Yes. Desired state converged, status condition with reason per pass Source for Levelrail, Level-triggered reconciler with status reasons
Kamal No. None: it runs Docker commands per deploy Source for Kamal, Level-triggered reconciler with status reasons
Single control plane binary
Levelrail Yes. One control plane binary, plus one agent per extra node Source for Levelrail, Single control plane binary
Kamal Not applicable
Observability
Metrics and log search built in
Levelrail Yes. Node-local metrics at 15s and full-text log search Source for Levelrail, Metrics and log search built in
Kamal No. Bring your own Source for Kamal, Metrics and log search built in
Crashloop diagnosis with last log lines
Levelrail Yes. Crashloop detection surfaces the last log lines Source for Levelrail, Crashloop diagnosis with last log lines
Kamal Not stated in their docs
Public status page
Levelrail Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS Source for Levelrail, Public status page
Kamal Not stated in their docs
Deploys
One-click rollback with pinned images
Levelrail Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta Source for Levelrail, One-click rollback with pinned images
Kamal Partly. Rollback to chosen version; old containers pruned after 3 days Source for Kamal, One-click rollback with pinned images
Readiness-gated deploys (zero-downtime on one node)
Levelrail Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta Source for Levelrail, Readiness-gated deploys (zero-downtime on one node)
Kamal Yes. Zero-downtime cutover with kamal-proxy Source for Kamal, Readiness-gated deploys (zero-downtime on one node)
Preview environments per pull request
Levelrail Yes. Stable on GitHub, beta on GitLab and Bitbucket Source for Levelrail, Preview environments per pull request
Kamal Not stated in their docs
Docker Compose apps
Levelrail Active. Pre-built images from the dashboard; git deploys can build Source for Levelrail, Docker Compose apps
Kamal Not stated in their docs
Git push deploys from GitHub, GitLab, Bitbucket
Levelrail Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta Source for Levelrail, Git push deploys from GitHub, GitLab, Bitbucket
Kamal Not stated in their docs
Builds from source without a Dockerfile
Levelrail Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected Source for Levelrail, Builds from source without a Dockerfile
Kamal Partly. Dockerfile required by default; buildpacks only when configured explicitly Source for Kamal, Builds from source without a Dockerfile
Scheduled jobs
Levelrail Yes. Per-app cron jobs with run history, logs, retries and alerts Source for Levelrail, Scheduled jobs
Kamal Not stated in their docs
Deploy freezes and policies
Levelrail Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys Source for Levelrail, Deploy freezes and policies
Kamal Partly. kamal lock acquire can block deploys during a maintenance window and a non-zero exit from pre-deploy or other hooks aborts the command (hooks can be skipped with --skip_hooks); no scheduled freeze windows or policy engine Source for Kamal, Deploy freezes and policies
CI deploys with no stored secret
Levelrail Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token Source for Levelrail, CI deploys with no stored secret
Kamal Not stated in their docs
Static site hosting
Levelrail Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy Source for Levelrail, Static site hosting
Kamal Not stated in their docs
Data
Managed databases with scheduled backups
Levelrail Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta Source for Levelrail, Managed databases with scheduled backups
Kamal Partly. Accessories run containers such as MySQL; no backups documented Source for Kamal, Managed databases with scheduled backups
Point-in-time database restore
Levelrail Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore Source for Levelrail, Point-in-time database restore
Kamal Not stated in their docs
Encrypted secrets for env vars
Levelrail Yes. Envelope-encrypted, masked in UI, injected at container create Source for Levelrail, Encrypted secrets for env vars
Kamal Partly. kamal secrets helpers extract secrets from password managers; custom TLS certs load from secrets; no encryption at rest documented Source for Kamal, Encrypted secrets for env vars
Networking
Private networking between apps or nodes
Levelrail Active. Named networks and allow rules; cross-server mesh unverified on real hosts Source for Levelrail, Private networking between apps or nodes
Kamal Partly. Containers share a custom Docker network per host, no cross server networking Source for Kamal, Private networking between apps or nodes
Load balancing across replicas
Levelrail Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only Source for Levelrail, Load balancing across replicas
Kamal No. Proxy config covers hosts, health checks, timeouts and buffering for one server; no balancing across servers is described Source for Kamal, Load balancing across replicas
Request firewall, rate limits, bot protection
Levelrail Yes. Ordered rules, IP bans, bot handling, challenges and rate limits Source for Levelrail, Request firewall, rate limits, bot protection
Kamal No. Proxy options cover buffering, body size limit and timeouts; no rate limits or bot protection described Source for Kamal, Request firewall, rate limits, bot protection
Automatic TLS certificates
Levelrail Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported Source for Levelrail, Automatic TLS certificates
Kamal Partly. kamal-proxy Let's Encrypt HTTPS, only with one server and host set Source for Kamal, Automatic TLS certificates
Platform
Multi-node
Levelrail Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta Source for Levelrail, Multi-node
Kamal Yes. SSH to each server Source for Kamal, Multi-node
One-click service templates
Levelrail Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet Source for Levelrail, One-click service templates
Kamal Not stated in their docs
Autoscaling
Levelrail Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons Source for Levelrail, Autoscaling
Kamal Not stated in their docs
Sleep idle apps, wake on request
Levelrail Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress Source for Levelrail, Sleep idle apps, wake on request
Kamal Not stated in their docs
CLI and HTTP API
Levelrail Yes. CLI and REST API documented; MCP server is beta Source for Levelrail, CLI and HTTP API
Kamal Not applicable
Deploy and alert notifications
Levelrail Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes Source for Levelrail, Deploy and alert notifications
Kamal Partly. Hooks run custom scripts; no built-in Slack, Discord or email Source for Kamal, Deploy and alert notifications
MCP / AI assistant integration
Levelrail Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag Source for Levelrail, MCP / AI assistant integration
Kamal Not stated in their docs
Runs on servers you own
Levelrail Yes. Runs on Linux servers you provide Source for Levelrail, Runs on servers you own
Kamal Yes. Runs against servers you own over SSH Source for Kamal, Runs on servers you own
Open source license
Levelrail Yes. Apache 2.0, one edition, no paid tier Source for Levelrail, Open source license
Kamal Yes. Released under the MIT License Source for Kamal, Open source license
Declarative apply of the whole platform
Levelrail Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag Source for Levelrail, Declarative apply of the whole platform
Kamal Partly. config/deploy.yml declares app, roles, proxy, accessories and servers and kamal setup/deploy applies it; commands run per deploy, no state reconciliation or drift detection documented Source for Kamal, Declarative apply of the whole platform
Functions and one-off jobs
Levelrail Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries Source for Levelrail, Functions and one-off jobs
Kamal Partly. A role can run a custom cmd such as bin/jobs for a job server; scheduled jobs are not described Source for Kamal, Functions and one-off jobs
Security
IAM, audit log and 2FA in the free edition
Levelrail Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition Source for Levelrail, IAM, audit log and 2FA in the free edition
Kamal Partly. Per-server audit log (.kamal/app-audit.log) of deploy steps, last 50 lines via kamal audit; no users, roles or 2FA Source for Kamal, IAM, audit log and 2FA in the free edition
SAML or SCIM single sign-on
Levelrail Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra Source for Levelrail, SAML or SCIM single sign-on
Kamal Not applicable
Audit log of every action
Levelrail Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket Source for Levelrail, Audit log of every action
Kamal Partly. Per-server .kamal/app-audit.log of deploy steps, read with kamal audit; entries carry a user tag but the page does not say which person; hooks expose KAMAL_PERFORMER for custom reports Source for Kamal, Audit log of every action
Two-person deploy approvals
Levelrail Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord Source for Levelrail, Two-person deploy approvals
Kamal Not stated in their docs
Architecture
No SSH: agents dial out
Levelrail Yes. Agents dial out over mTLS; no inbound ports Source for Levelrail, No SSH: agents dial out
Vercel Not applicable
Level-triggered reconciler with status reasons
Levelrail Yes. Desired state converged, status condition with reason per pass Source for Levelrail, Level-triggered reconciler with status reasons
Vercel Not applicable
Single control plane binary
Levelrail Yes. One control plane binary, plus one agent per extra node Source for Levelrail, Single control plane binary
Vercel Not applicable
Observability
Metrics and log search built in
Levelrail Yes. Node-local metrics at 15s and full-text log search Source for Levelrail, Metrics and log search built in
Vercel Partly. Observability on all plans; Observability Plus ($1.20 per 1M events) adds retention and more metrics; runtime log retention on Enterprise is 3 days, 30 with Plus; CLI log query and request tracing Source for Vercel, Metrics and log search built in
Crashloop diagnosis with last log lines
Levelrail Yes. Crashloop detection surfaces the last log lines Source for Levelrail, Crashloop diagnosis with last log lines
Vercel Not stated in their docs
Public status page
Levelrail Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS Source for Levelrail, Public status page
Vercel Not stated in their docs
Deploys
One-click rollback with pinned images
Levelrail Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta Source for Levelrail, One-click rollback with pinned images
Vercel Yes. Instant Rollback repoints production domains to an earlier production deployment with no rebuild. Hobby can only roll back to the immediately previous deployment; Pro and Enterprise can pick any eligible one. After a rollback auto-assignment of production domains is turned off until you promote a deployment Source for Vercel, One-click rollback with pinned images
Readiness-gated deploys (zero-downtime on one node)
Levelrail Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta Source for Levelrail, Readiness-gated deploys (zero-downtime on one node)
Vercel Not stated in their docs
Preview environments per pull request
Levelrail Yes. Stable on GitHub, beta on GitLab and Bitbucket Source for Levelrail, Preview environments per pull request
Vercel Yes. Preview deployment with unique branch and commit URLs for every non-production push, PR (GitHub, GitLab, Bitbucket) or CLI deploy. Protected by Vercel Authentication by default; password protection is $20 per project per month on Pro; custom preview suffix $100 per month on Pro Source for Vercel, Preview environments per pull request
Docker Compose apps
Levelrail Active. Pre-built images from the dashboard; git deploys can build Source for Levelrail, Docker Compose apps
Vercel No. Compose file is not read ('Vercel does not read any of it'). Services are declared in vercel.json (Vercel Services and container images are both in beta on all plans); migration is manual and some patterns (bind mounts, depends_on ordering, privileged containers) have no equivalent Source for Vercel, Docker Compose apps
Git push deploys from GitHub, GitLab, Bitbucket
Levelrail Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta Source for Levelrail, Git push deploys from GitHub, GitLab, Bitbucket
Vercel Yes. Auto deploys on push from GitHub, GitLab and Bitbucket Source for Vercel, Git push deploys from GitHub, GitLab, Bitbucket
Builds from source without a Dockerfile
Levelrail Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected Source for Levelrail, Builds from source without a Dockerfile
Vercel Yes. Framework auto-detection builds with no Dockerfile needed Source for Vercel, Builds from source without a Dockerfile
Scheduled jobs
Levelrail Yes. Per-app cron jobs with run history, logs, retries and alerts Source for Levelrail, Scheduled jobs
Vercel Yes. Cron Jobs on all plans (UTC, run Vercel Functions); docs now mark Cron Jobs as legacy, moving to Schedules with timezones and runtime creation Source for Vercel, Scheduled jobs
Deploy freezes and policies
Levelrail Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys Source for Levelrail, Deploy freezes and policies
Vercel Partly. Deployment Checks (native lint and typecheck, GitHub Actions results, Marketplace checks) hold production promotion until they pass; no freeze windows stated; Force Promote bypasses Source for Vercel, Deploy freezes and policies
CI deploys with no stored secret
Levelrail Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token Source for Levelrail, CI deploys with no stored secret
Vercel Not stated in their docs
Static site hosting
Levelrail Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy Source for Levelrail, Static site hosting
Vercel Yes. Framework auto-detected builds served from the CDN; static assets are counted as CDN requests and Hobby includes 1M of them Source for Vercel, Static site hosting
Data
Managed databases with scheduled backups
Levelrail Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta Source for Levelrail, Managed databases with scheduled backups
Vercel Partly. Marketplace databases from the dashboard; backup schedules not stated Source for Vercel, Managed databases with scheduled backups
Point-in-time database restore
Levelrail Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore Source for Levelrail, Point-in-time database restore
Vercel Not stated in their docs
Encrypted secrets for env vars
Levelrail Yes. Envelope-encrypted, masked in UI, injected at container create Source for Levelrail, Encrypted secrets for env vars
Vercel Yes. Encrypted at rest; Secret type values are write-only after saving Source for Vercel, Encrypted secrets for env vars
Networking
Private networking between apps or nodes
Levelrail Active. Named networks and allow rules; cross-server mesh unverified on real hosts Source for Levelrail, Private networking between apps or nodes
Vercel Partly. Secure Compute private networks and VPC peering, Enterprise only Source for Vercel, Private networking between apps or nodes
Load balancing across replicas
Levelrail Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only Source for Levelrail, Load balancing across replicas
Vercel Not applicable
Request firewall, rate limits, bot protection
Levelrail Yes. Ordered rules, IP bans, bot handling, challenges and rate limits Source for Levelrail, Request firewall, rate limits, bot protection
Vercel Partly. DDoS mitigation on all plans; WAF on all plans with 3 custom rules on Hobby, 40 on Pro, 250 on Enterprise; Bot Protection and managed rulesets; managed OWASP ruleset is Enterprise and custom priced Source for Vercel, Request firewall, rate limits, bot protection
Automatic TLS certificates
Levelrail Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported Source for Levelrail, Automatic TLS certificates
Vercel Yes. Let's Encrypt certs issued and renewed automatically per domain Source for Vercel, Automatic TLS certificates
Platform
Multi-node
Levelrail Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta Source for Levelrail, Multi-node
Vercel Not applicable
One-click service templates
Levelrail Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet Source for Levelrail, One-click service templates
Vercel Partly. Framework starter templates listed; no service template catalog stated Source for Vercel, One-click service templates
Autoscaling
Levelrail Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons Source for Levelrail, Autoscaling
Vercel Yes. Fluid compute scales instances automatically with traffic Source for Vercel, Autoscaling
Sleep idle apps, wake on request
Levelrail Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress Source for Levelrail, Sleep idle apps, wake on request
Vercel Not stated in their docs
CLI and HTTP API
Levelrail Yes. CLI and REST API documented; MCP server is beta Source for Levelrail, CLI and HTTP API
Vercel Yes. CLI covers deploy, logs, env, promote; REST API also available Source for Vercel, CLI and HTTP API
Deploy and alert notifications
Levelrail Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes Source for Levelrail, Deploy and alert notifications
Vercel Partly. Deployment failure notices by dashboard, email and push on all plans; anomaly alerts by email, Slack or webhook, but those need Pro or Enterprise with Observability Plus Source for Vercel, Deploy and alert notifications
MCP / AI assistant integration
Levelrail Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag Source for Levelrail, MCP / AI assistant integration
Vercel Yes. Official remote Vercel MCP server at mcp.vercel.com, OAuth Source for Vercel, MCP / AI assistant integration
Runs on servers you own
Levelrail Yes. Runs on Linux servers you provide Source for Levelrail, Runs on servers you own
Vercel Not applicable
Open source license
Levelrail Yes. Apache 2.0, one edition, no paid tier Source for Levelrail, Open source license
Vercel Not stated in their docs
Declarative apply of the whole platform
Levelrail Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag Source for Levelrail, Declarative apply of the whole platform
Vercel Partly. Per-project config in vercel.json or vercel.ts; WAF can be managed through REST API, SDK or Terraform; no whole-platform declarative apply stated Source for Vercel, Declarative apply of the whole platform
Functions and one-off jobs
Levelrail Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries Source for Levelrail, Functions and one-off jobs
Vercel Yes. Vercel Functions on Fluid compute, Cron Jobs on all plans, plus Queues, Workflow and Sandbox listed as background options Source for Vercel, Functions and one-off jobs
Security
IAM, audit log and 2FA in the free edition
Levelrail Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition Source for Levelrail, IAM, audit log and 2FA in the free edition
Vercel Partly. Team roles on Pro and Enterprise; project-level roles, Security and Contributor roles and audit logs are Enterprise only; no audit logs on Hobby or Pro Source for Vercel, IAM, audit log and 2FA in the free edition
SAML or SCIM single sign-on
Levelrail Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra Source for Levelrail, SAML or SCIM single sign-on
Vercel Partly. SAML SSO on Pro as a $300/month add-on and included on Enterprise; Directory Sync (SCIM) on Enterprise only Source for Vercel, SAML or SCIM single sign-on
Audit log of every action
Levelrail Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket Source for Levelrail, Audit log of every action
Vercel Partly. Audit Logs are Enterprise only, owner and security roles; CSV export covers the last 90 days. Custom SIEM streaming is being replaced by Audit Log Drains Source for Vercel, Audit log of every action
Two-person deploy approvals
Levelrail Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord Source for Levelrail, Two-person deploy approvals
Vercel Not stated in their docs
Architecture
No SSH: agents dial out
Levelrail Yes. Agents dial out over mTLS; no inbound ports Source for Levelrail, No SSH: agents dial out
Heroku Not applicable
Level-triggered reconciler with status reasons
Levelrail Yes. Desired state converged, status condition with reason per pass Source for Levelrail, Level-triggered reconciler with status reasons
Heroku Not applicable
Single control plane binary
Levelrail Yes. One control plane binary, plus one agent per extra node Source for Levelrail, Single control plane binary
Heroku Not applicable
Observability
Metrics and log search built in
Levelrail Yes. Node-local metrics at 15s and full-text log search Source for Levelrail, Metrics and log search built in
Heroku Partly. Dyno metrics and events shown; log search not documented Source for Heroku, Metrics and log search built in
Crashloop diagnosis with last log lines
Levelrail Yes. Crashloop detection surfaces the last log lines Source for Levelrail, Crashloop diagnosis with last log lines
Heroku Partly. Error codes explain crashes in logs; no automatic diagnosis documented Source for Heroku, Crashloop diagnosis with last log lines
Public status page
Levelrail Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS Source for Levelrail, Public status page
Heroku Not stated in their docs
Deploys
One-click rollback with pinned images
Levelrail Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta Source for Levelrail, One-click rollback with pinned images
Heroku Yes. heroku rollback creates a release reusing the earlier slug or image Source for Heroku, One-click rollback with pinned images
Readiness-gated deploys (zero-downtime on one node)
Levelrail Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta Source for Levelrail, Readiness-gated deploys (zero-downtime on one node)
Heroku Partly. Preboot is opt-in, standard and performance dynos; not guaranteed Source for Heroku, Readiness-gated deploys (zero-downtime on one node)
Preview environments per pull request
Levelrail Yes. Stable on GitHub, beta on GitLab and Bitbucket Source for Levelrail, Preview environments per pull request
Heroku Yes. Review apps per GitHub pull request; GitHub only Source for Heroku, Preview environments per pull request
Docker Compose apps
Levelrail Active. Pre-built images from the dashboard; git deploys can build Source for Levelrail, Docker Compose apps
Heroku No. Docker Compose is mentioned for local development only; deploy options are container:push and release, or heroku.yml builds (heroku.yml Docker builds are unsupported on Fir) Source for Heroku, Docker Compose apps
Git push deploys from GitHub, GitLab, Bitbucket
Levelrail Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta Source for Levelrail, Git push deploys from GitHub, GitLab, Bitbucket
Heroku Partly. GitHub automatic deploys (GitHub Enterprise Server and Cloud for Enterprise customers); GitLab and Bitbucket go through CI with dpl, git hooks, the Platform API or Terraform Source for Heroku, Git push deploys from GitHub, GitLab, Bitbucket
Builds from source without a Dockerfile
Levelrail Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected Source for Levelrail, Builds from source without a Dockerfile
Heroku Yes. Buildpacks auto-detect language and build source with no Dockerfile Source for Heroku, Builds from source without a Dockerfile
Scheduled jobs
Levelrail Yes. Per-app cron jobs with run history, logs, retries and alerts Source for Levelrail, Scheduled jobs
Heroku Partly. Heroku Scheduler is a free add-on running every 10 minutes, hourly or daily (daily in UTC); jobs bill as one-off dyno hours; Heroku advises a custom clock process for critical jobs because runs can be skipped or duplicated Source for Heroku, Scheduled jobs
Deploy freezes and policies
Levelrail Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys Source for Levelrail, Deploy freezes and policies
Heroku Partly. 'Wait for CI to pass before deploy' holds GitHub auto-deploys until commit statuses succeed; no freeze windows stated Source for Heroku, Deploy freezes and policies
CI deploys with no stored secret
Levelrail Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token Source for Levelrail, CI deploys with no stored secret
Heroku Not stated in their docs
Static site hosting
Levelrail Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy Source for Levelrail, Static site hosting
Heroku Not stated in their docs
Data
Managed databases with scheduled backups
Levelrail Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta Source for Levelrail, Managed databases with scheduled backups
Heroku Partly. Postgres Essential-0 $5 (1 GB, 20 connections, no rollback, no fork/follow, 99.5% uptime); daily logical backups; Advanced tier (usage-based, from $150, HA by default) replaces Classic and Heroku PGBackups is not supported on Advanced Source for Heroku, Managed databases with scheduled backups
Point-in-time database restore
Levelrail Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore Source for Levelrail, Point-in-time database restore
Heroku Partly. Rollback window by tier: Standard 4 days, Premium/Private/Shield/Advanced 7 days; Essential has no rollback Source for Heroku, Point-in-time database restore
Encrypted secrets for env vars
Levelrail Yes. Envelope-encrypted, masked in UI, injected at container create Source for Levelrail, Encrypted secrets for env vars
Heroku Not stated in their docs
Networking
Private networking between apps or nodes
Levelrail Active. Named networks and allow rules; cross-server mesh unverified on real hosts Source for Levelrail, Private networking between apps or nodes
Heroku Partly. Private Spaces, Teams or Enterprise only: internal routing and DNS Source for Heroku, Private networking between apps or nodes
Load balancing across replicas
Levelrail Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only Source for Levelrail, Load balancing across replicas
Heroku Yes. Router randomly balances requests across web dynos Source for Heroku, Load balancing across replicas
Request firewall, rate limits, bot protection
Levelrail Yes. Ordered rules, IP bans, bot handling, challenges and rate limits Source for Levelrail, Request firewall, rate limits, bot protection
Heroku Not stated in their docs
Automatic TLS certificates
Levelrail Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported Source for Levelrail, Automatic TLS certificates
Heroku Yes. Automated Certificate Management: free Let's Encrypt certs for custom domains Source for Heroku, Automatic TLS certificates
Platform
Multi-node
Levelrail Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta Source for Levelrail, Multi-node
Heroku Not applicable
One-click service templates
Levelrail Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet Source for Levelrail, One-click service templates
Heroku Partly. app.json and Deploy button; no one-click service catalog Source for Heroku, One-click service templates
Autoscaling
Levelrail Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons Source for Levelrail, Autoscaling
Heroku Partly. Web dynos only, Performance, Private or Shield tiers, scaled on a p95 response time target; not available for Fir apps Source for Heroku, Autoscaling
Sleep idle apps, wake on request
Levelrail Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress Source for Levelrail, Sleep idle apps, wake on request
Heroku Partly. Eco web dynos sleep after 30 minutes without web traffic; other dyno types do not Source for Heroku, Sleep idle apps, wake on request
CLI and HTTP API
Levelrail Yes. CLI and REST API documented; MCP server is beta Source for Levelrail, CLI and HTTP API
Heroku Partly. CLI and Platform API for apps and config; full parity not stated Source for Heroku, CLI and HTTP API
Deploy and alert notifications
Levelrail Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes Source for Levelrail, Deploy and alert notifications
Heroku Partly. Threshold alerts on p95 response time or 5xx rate by email, PagerDuty or dashboard; Slack through ChatOps; webhooks via Platform API Source for Heroku, Deploy and alert notifications
MCP / AI assistant integration
Levelrail Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag Source for Levelrail, MCP / AI assistant integration
Heroku Yes. Remote MCP server at mcp.heroku.com with OAuth login Source for Heroku, MCP / AI assistant integration
Runs on servers you own
Levelrail Yes. Runs on Linux servers you provide Source for Levelrail, Runs on servers you own
Heroku Not applicable
Open source license
Levelrail Yes. Apache 2.0, one edition, no paid tier Source for Levelrail, Open source license
Heroku Not stated in their docs
Declarative apply of the whole platform
Levelrail Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag Source for Levelrail, Declarative apply of the whole platform
Heroku Partly. Terraform listed as a deploy path and app.json for app setup; Platform API manages apps, pipelines, config vars and add-ons; no whole-platform declarative apply stated Source for Heroku, Declarative apply of the whole platform
Functions and one-off jobs
Levelrail Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries Source for Levelrail, Functions and one-off jobs
Heroku Partly. Worker dynos, one-off dynos and the free Scheduler add-on (10 minute, hourly or daily); no function-style runtime stated Source for Heroku, Functions and one-off jobs
Security
IAM, audit log and 2FA in the free edition
Levelrail Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition Source for Levelrail, IAM, audit log and 2FA in the free edition
Heroku Partly. Four team roles and mandatory MFA; audit trails Enterprise only Source for Heroku, IAM, audit log and 2FA in the free edition
SAML or SCIM single sign-on
Levelrail Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra Source for Levelrail, SAML or SCIM single sign-on
Heroku Partly. SAML SSO for Heroku Teams and Enterprise; accounts auto-provisioned on first IdP login, removal needs Support; SCIM not documented Source for Heroku, SAML or SCIM single sign-on
Audit log of every action
Levelrail Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket Source for Levelrail, Audit log of every action
Heroku Partly. Audit trail export (JSON archive per month) for Enterprise accounts only, needs manage permission Source for Heroku, Audit log of every action
Two-person deploy approvals
Levelrail Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord Source for Levelrail, Two-person deploy approvals
Heroku Not stated in their docs
Architecture
No SSH: agents dial out
Levelrail Yes. Agents dial out over mTLS; no inbound ports Source for Levelrail, No SSH: agents dial out
Railway Not applicable
Level-triggered reconciler with status reasons
Levelrail Yes. Desired state converged, status condition with reason per pass Source for Levelrail, Level-triggered reconciler with status reasons
Railway Not applicable
Single control plane binary
Levelrail Yes. One control plane binary, plus one agent per extra node Source for Levelrail, Single control plane binary
Railway Not applicable
Observability
Metrics and log search built in
Levelrail Yes. Node-local metrics at 15s and full-text log search Source for Levelrail, Metrics and log search built in
Railway Yes. Dashboard CPU, memory, network, disk metrics; log search Source for Railway, Metrics and log search built in
Crashloop diagnosis with last log lines
Levelrail Yes. Crashloop detection surfaces the last log lines Source for Levelrail, Crashloop diagnosis with last log lines
Railway Partly. Default On Failure restart policy restarts up to 10 times; Free and Trial capped at 10 and cannot use Always; paid plans have no restart limit. No automatic diagnosis Source for Railway, Crashloop diagnosis with last log lines
Public status page
Levelrail Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS Source for Levelrail, Public status page
Railway Not stated in their docs
Deploys
One-click rollback with pinned images
Levelrail Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta Source for Levelrail, One-click rollback with pinned images
Railway Partly. Dashboard rollback to prior image; images kept 24 to 360 hours Source for Railway, One-click rollback with pinned images
Readiness-gated deploys (zero-downtime on one node)
Levelrail Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta Source for Levelrail, Readiness-gated deploys (zero-downtime on one node)
Railway Partly. Healthcheck gates cutover; volume services see brief downtime Source for Railway, Readiness-gated deploys (zero-downtime on one node)
Preview environments per pull request
Levelrail Yes. Stable on GitHub, beta on GitLab and Bitbucket Source for Levelrail, Preview environments per pull request
Railway Yes. Environments per PR Source for Railway, Preview environments per pull request
Docker Compose apps
Levelrail Active. Pre-built images from the dashboard; git deploys can build Source for Levelrail, Docker Compose apps
Railway Partly. Imports Compose services; docker-compose.yml not run directly Source for Railway, Docker Compose apps
Git push deploys from GitHub, GitLab, Bitbucket
Levelrail Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta Source for Levelrail, Git push deploys from GitHub, GitLab, Bitbucket
Railway Partly. Autodeploy from GitHub only; GitLab and Bitbucket not documented Source for Railway, Git push deploys from GitHub, GitLab, Bitbucket
Builds from source without a Dockerfile
Levelrail Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected Source for Levelrail, Builds from source without a Dockerfile
Railway Yes. Railpack detects stack without Dockerfile; Dockerfile takes priority Source for Railway, Builds from source without a Dockerfile
Scheduled jobs
Levelrail Yes. Per-app cron jobs with run history, logs, retries and alerts Source for Levelrail, Scheduled jobs
Railway Yes. Cron schedule per service; runs at least 5 minutes apart, UTC Source for Railway, Scheduled jobs
Deploy freezes and policies
Levelrail Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys Source for Levelrail, Deploy freezes and policies
Railway Partly. Wait for CI holds deploys until GitHub Actions check suites finish; Enterprise guardrails restrict deploy sources; no freeze windows documented Source for Railway, Deploy freezes and policies
CI deploys with no stored secret
Levelrail Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token Source for Levelrail, CI deploys with no stored secret
Railway No. No OIDC or keyless CI deploy documented; deploys use project or account tokens stored as CI secrets Source for Railway, CI deploys with no stored secret
Static site hosting
Levelrail Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy Source for Levelrail, Static site hosting
Railway Yes. Zero-config static site deploys from GitHub with custom domains and automatic SSL Source for Railway, Static site hosting
Data
Managed databases with scheduled backups
Levelrail Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta Source for Levelrail, Managed databases with scheduled backups
Railway Partly. Scheduled volume and Postgres PITR backups; S3 export needs own job Source for Railway, Managed databases with scheduled backups
Point-in-time database restore
Levelrail Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore Source for Levelrail, Point-in-time database restore
Railway Not stated in their docs
Encrypted secrets for env vars
Levelrail Yes. Envelope-encrypted, masked in UI, injected at container create Source for Levelrail, Encrypted secrets for env vars
Railway Yes. Variables encrypted at rest; sealed values write-only Source for Railway, Encrypted secrets for env vars
Networking
Private networking between apps or nodes
Levelrail Active. Named networks and allow rules; cross-server mesh unverified on real hosts Source for Levelrail, Private networking between apps or nodes
Railway Yes. WireGuard private DNS per environment, service.railway.internal Source for Railway, Private networking between apps or nodes
Load balancing across replicas
Levelrail Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only Source for Levelrail, Load balancing across replicas
Railway Yes. Traffic spread across replicas and regions; no sticky sessions Source for Railway, Load balancing across replicas
Request firewall, rate limits, bot protection
Levelrail Yes. Ordered rules, IP bans, bot handling, challenges and rate limits Source for Levelrail, Request firewall, rate limits, bot protection
Railway Partly. Under Attack Mode plus Edge rules (block, allow, challenge, redirect, cache override); rate limiting not documented Source for Railway, Request firewall, rate limits, bot protection
Automatic TLS certificates
Levelrail Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported Source for Levelrail, Automatic TLS certificates
Railway Yes. Let's Encrypt certificates for custom domains, renewed automatically Source for Railway, Automatic TLS certificates
Platform
Multi-node
Levelrail Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta Source for Levelrail, Multi-node
Railway Not applicable
One-click service templates
Levelrail Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet Source for Levelrail, One-click service templates
Railway Yes. Template marketplace with one-click deploys Source for Railway, One-click service templates
Autoscaling
Levelrail Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons Source for Levelrail, Autoscaling
Railway No. Vertical scaling is automatic per container, but replica count stays where you set it; horizontal autoscaling means building your own with the API, railway scale CLI or cron Source for Railway, Autoscaling
Sleep idle apps, wake on request
Levelrail Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress Source for Levelrail, Sleep idle apps, wake on request
Railway Yes. Serverless setting per service; sleeps 5 to 10 minutes after last outbound traffic, wakes on inbound request, first request may be slow or return 502 Source for Railway, Sleep idle apps, wake on request
CLI and HTTP API
Levelrail Yes. CLI and REST API documented; MCP server is beta Source for Levelrail, CLI and HTTP API
Railway Yes. Public GraphQL API is the same API that powers the dashboard; CLI and MCP server also available Source for Railway, CLI and HTTP API
Deploy and alert notifications
Levelrail Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes Source for Levelrail, Deploy and alert notifications
Railway Yes. Slack and Discord webhooks; email via account notifications Source for Railway, Deploy and alert notifications
MCP / AI assistant integration
Levelrail Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag Source for Levelrail, MCP / AI assistant integration
Railway Yes. Official hosted MCP server at mcp.railway.com Source for Railway, MCP / AI assistant integration
Runs on servers you own
Levelrail Yes. Runs on Linux servers you provide Source for Levelrail, Runs on servers you own
Railway Not applicable
Open source license
Levelrail Yes. Apache 2.0, one edition, no paid tier Source for Levelrail, Open source license
Railway Not stated in their docs
Declarative apply of the whole platform
Levelrail Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag Source for Levelrail, Declarative apply of the whole platform
Railway Yes. Infrastructure as Code in .railway/railway.ts (GA; Python and Go in beta) defines services, databases, volumes, buckets, domains, variables, replicas; railway config plan and apply; railway.json/toml Config as Code is deprecated, legacy files read until 2026-12-01 Source for Railway, Declarative apply of the whole platform
Functions and one-off jobs
Levelrail Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries Source for Levelrail, Functions and one-off jobs
Railway Yes. Cron services plus Functions (single TypeScript file on Bun, with optional cron schedule) Source for Railway, Functions and one-off jobs
Security
IAM, audit log and 2FA in the free edition
Levelrail Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition Source for Levelrail, IAM, audit log and 2FA in the free edition
Railway Partly. Audit logs: 48 hours on Free, Trial and Hobby, 30 days on Pro, 18 months on Enterprise; Environment RBAC and SAML SSO are Enterprise or committed-spend only Source for Railway, IAM, audit log and 2FA in the free edition
SAML or SCIM single sign-on
Levelrail Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra Source for Levelrail, SAML or SCIM single sign-on
Railway Partly. SAML SSO is an Enterprise feature; SCIM not mentioned Source for Railway, SAML or SCIM single sign-on
Audit log of every action
Levelrail Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket Source for Levelrail, Audit log of every action
Railway Yes. Retention 48 hours on Free, Trial and Hobby, 30 days on Pro, 18 months on Enterprise; viewable by workspace admins Source for Railway, Audit log of every action
Two-person deploy approvals
Levelrail Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord Source for Levelrail, Two-person deploy approvals
Railway Partly. Deployment Approvals hold commits from users without a linked Railway account; duplicated environments need staged-change review; Environment RBAC and approved-sources guardrail are Enterprise or committed-spend; no two-person production approval documented Source for Railway, Two-person deploy approvals
Architecture
No SSH: agents dial out
Levelrail Yes. Agents dial out over mTLS; no inbound ports Source for Levelrail, No SSH: agents dial out
Cloud Run Not applicable
Level-triggered reconciler with status reasons
Levelrail Yes. Desired state converged, status condition with reason per pass Source for Levelrail, Level-triggered reconciler with status reasons
Cloud Run Not applicable
Single control plane binary
Levelrail Yes. One control plane binary, plus one agent per extra node Source for Levelrail, Single control plane binary
Cloud Run Not applicable
Observability
Metrics and log search built in
Levelrail Yes. Node-local metrics at 15s and full-text log search Source for Levelrail, Metrics and log search built in
Cloud Run Yes. Automatic Cloud Monitoring metrics; logs in Cloud Logging Source for Cloud Run, Metrics and log search built in
Crashloop diagnosis with last log lines
Levelrail Yes. Crashloop detection surfaces the last log lines Source for Levelrail, Crashloop diagnosis with last log lines
Cloud Run Partly. Troubleshooting guide; logs checked manually, no automatic crashloop view Source for Cloud Run, Crashloop diagnosis with last log lines
Public status page
Levelrail Yes. Opt-in public page with components, 90 day uptime bars, incidents, maintenance and RSS Source for Levelrail, Public status page
Cloud Run Not stated in their docs
Deploys
One-click rollback with pinned images
Levelrail Yes. Prior images pinned so cleanup cannot remove them. A 5xx spike or crashloop after cutover rolls back automatically (live-tested); the latency signal is beta Source for Levelrail, One-click rollback with pinned images
Cloud Run Yes. Traffic rolls back to a previous revision immediately Source for Cloud Run, One-click rollback with pinned images
Readiness-gated deploys (zero-downtime on one node)
Levelrail Yes. Cutover only after the readiness probe passes. Live-tested on one node; multi-node is beta Source for Levelrail, Readiness-gated deploys (zero-downtime on one node)
Cloud Run Partly. Gradual rollouts and traffic splits; health gating of cutover not stated Source for Cloud Run, Readiness-gated deploys (zero-downtime on one node)
Preview environments per pull request
Levelrail Yes. Stable on GitHub, beta on GitLab and Bitbucket Source for Levelrail, Preview environments per pull request
Cloud Run Partly. Revision tags give separate URLs without traffic; no PR previews Source for Cloud Run, Preview environments per pull request
Docker Compose apps
Levelrail Active. Pre-built images from the dashboard; git deploys can build Source for Levelrail, Docker Compose apps
Cloud Run Partly. gcloud run compose up deploys a Compose file as one multi container service (supported subset of fields, secrets map to Secret Manager, volumes to Cloud Storage, max instances 1 by default); Google recommends Terraform for production Source for Cloud Run, Docker Compose apps
Git push deploys from GitHub, GitLab, Bitbucket
Levelrail Active. GitHub, GitLab, Bitbucket and Gitea documented; GitLab and Bitbucket previews are beta (synthetic webhooks). A git push remote on your own server (no provider needed) is beta Source for Levelrail, Git push deploys from GitHub, GitLab, Bitbucket
Cloud Run Yes. Developer Connect covers GitHub, GitLab, Bitbucket; Cloud Build covers GitHub Source for Cloud Run, Git push deploys from GitHub, GitLab, Bitbucket
Builds from source without a Dockerfile
Levelrail Active. Railpack detects Node, Python, Go and more. Ruby and PHP are accepted but only plan generation is verified, no live build; Rust, Deno and Elixir are rejected Source for Levelrail, Builds from source without a Dockerfile
Cloud Run Yes. gcloud run deploy --source uses a Dockerfile if present, otherwise Google Cloud buildpacks (Go, Node.js, Python, Java, .NET, Ruby, PHP); deploy-without-build is Preview Source for Cloud Run, Builds from source without a Dockerfile
Scheduled jobs
Levelrail Yes. Per-app cron jobs with run history, logs, retries and alerts Source for Levelrail, Scheduled jobs
Cloud Run Partly. Jobs can run once, on a schedule or in a workflow; delayed job execution is in Preview Source for Cloud Run, Scheduled jobs
Deploy freezes and policies
Levelrail Yes. Cron-defined freeze windows per app with an audited override reason; also gate image auto-update and scheduled deploys Source for Levelrail, Deploy freezes and policies
Cloud Run Partly. Binary Authorization policy lets Cloud Run deploy only images signed by trusted authorities, with allowlist patterns; freeze windows not stated Source for Cloud Run, Deploy freezes and policies
CI deploys with no stored secret
Levelrail Active. GitHub Actions only: a GitHub OIDC token is exchanged for a 15 minute token scoped to one app. Other CI uses an API token Source for Levelrail, CI deploys with no stored secret
Cloud Run Yes. Workload Identity Federation lets GitHub Actions and GitLab CI jobs use their OIDC tokens for Google Cloud credentials, without service account keys Source for Cloud Run, CI deploys with no stored secret
Static site hosting
Levelrail Yes. Static HTML from git served by the embedded ingress with TLS; drag-and-drop folder or zip deploy Source for Levelrail, Static site hosting
Cloud Run Not stated in their docs
Data
Managed databases with scheduled backups
Levelrail Active. Many engines, scheduled S3 backups. Postgres point-in-time restore is tested live; MySQL, MariaDB and MongoDB are beta Source for Levelrail, Managed databases with scheduled backups
Cloud Run Not stated in their docs
Point-in-time database restore
Levelrail Active. Postgres stable with an S3 compatible bucket; MySQL and MariaDB beta; MongoDB has no point-in-time restore Source for Levelrail, Point-in-time database restore
Cloud Run Not stated in their docs
Encrypted secrets for env vars
Levelrail Yes. Envelope-encrypted, masked in UI, injected at container create Source for Levelrail, Encrypted secrets for env vars
Cloud Run Yes. Secret Manager values injected as env vars or mounted volumes Source for Cloud Run, Encrypted secrets for env vars
Networking
Private networking between apps or nodes
Levelrail Active. Named networks and allow rules; cross-server mesh unverified on real hosts Source for Levelrail, Private networking between apps or nodes
Cloud Run Partly. Direct VPC egress reaches private IPs; internal DNS not documented Source for Cloud Run, Private networking between apps or nodes
Load balancing across replicas
Levelrail Active. Apps with more than one replica are round-robin balanced by default, with failed-replica skip and retry. The configurable balancer (health checks, sticky sessions, weights, drain) is experimental, behind a flag, verified on one machine only. HTTP only Source for Levelrail, Load balancing across replicas
Cloud Run Yes. Requests balanced across regional instances automatically; multi region needs load balancer Source for Cloud Run, Load balancing across replicas
Request firewall, rate limits, bot protection
Levelrail Yes. Ordered rules, IP bans, bot handling, challenges and rate limits Source for Levelrail, Request firewall, rate limits, bot protection
Cloud Run Partly. Cloud Armor filters traffic only through an external load balancer Source for Cloud Run, Request firewall, rate limits, bot protection
Automatic TLS certificates
Levelrail Active. Self-signed by default; public ACME (HTTP-01) has one recorded run on a live VPS. Renewal near expiry and DNS-01 wildcards are not exercised live; CAs needing EAB are unsupported Source for Levelrail, Automatic TLS certificates
Cloud Run Partly. Domain mapping gives auto-issued and renewed Google-managed certs but is Preview and not recommended for production; global external Application Load Balancer is the recommended path. New Preview (Sep 28) custom URLs like example.cloud.run need no DNS or load balancer Source for Cloud Run, Automatic TLS certificates
Platform
Multi-node
Levelrail Active. Join, cordon and drain with real container relocation verified across two local Docker daemons; WireGuard mesh and cross-host transport still unverified. Resource-aware placement is beta Source for Levelrail, Multi-node
Cloud Run Not applicable
One-click service templates
Levelrail Yes. 592 catalog entries, 583 active; 346 proven on a real engine, 246 previews without that pass yet Source for Levelrail, One-click service templates
Cloud Run Not stated in their docs
Autoscaling
Levelrail Active. Beta: CPU, memory or request-rate scaling with cooldowns and reasons Source for Levelrail, Autoscaling
Cloud Run Yes. Scales on requests, events and CPU use; min and max instances; manual scaling available; custom CPU or concurrency utilization targets reached GA on 2026-09-29 Source for Cloud Run, Autoscaling
Sleep idle apps, wake on request
Levelrail Yes. Opt in per app; stops after 5 minutes to 7 days idle and wakes on the next request. Verified live against a real Docker daemon and the real ingress Source for Levelrail, Sleep idle apps, wake on request
Cloud Run Yes. Scales to zero instances by default when there is no traffic Source for Cloud Run, Sleep idle apps, wake on request
CLI and HTTP API
Levelrail Yes. CLI and REST API documented; MCP server is beta Source for Levelrail, CLI and HTTP API
Cloud Run Yes. gcloud CLI and Admin API manage services and jobs Source for Cloud Run, CLI and HTTP API
Deploy and alert notifications
Levelrail Active. Webhook and SMTP email have live e2e; 15 vendor kinds (Slack, Discord, PagerDuty and more) are beta, tested against local fakes Source for Levelrail, Deploy and alert notifications
Cloud Run Partly. Cloud Monitoring alerts to email, Slack, webhooks, Pub/Sub Source for Cloud Run, Deploy and alert notifications
MCP / AI assistant integration
Levelrail Yes. MCP server (beta) with over 240 tools in toolsets an agent enables on demand (256 at last count, 21 in the default session); in-app chat is behind a flag Source for Levelrail, MCP / AI assistant integration
Cloud Run Partly. Remote MCP server (Streamable HTTP, OAuth 2.0 and IAM) deploys from images, source or zip and lists and inspects services; regional endpoints and the /deploy prompt are Preview; Cloud Audit Logs and optional Model Armor Source for Cloud Run, MCP / AI assistant integration
Runs on servers you own
Levelrail Yes. Runs on Linux servers you provide Source for Levelrail, Runs on servers you own
Cloud Run Not applicable
Open source license
Levelrail Yes. Apache 2.0, one edition, no paid tier Source for Levelrail, Open source license
Cloud Run Not stated in their docs
Declarative apply of the whole platform
Levelrail Active. YAML export, plan and apply from CLI, dashboard, MCP or CI, plus a Terraform and OpenTofu provider. Experimental, behind a flag Source for Levelrail, Declarative apply of the whole platform
Cloud Run Partly. Per service Knative YAML export and 'gcloud run services replace', plus Terraform google_cloud_run_v2_service; whole-platform apply is Terraform across Google Cloud Source for Cloud Run, Declarative apply of the whole platform
Functions and one-off jobs
Levelrail Active. Node, Python or Go handlers that sleep when idle and wake on the first request; scheduled and background calls with retries Source for Levelrail, Functions and one-off jobs
Cloud Run Yes. Services plus Cloud Run jobs (up to 10,000 parallel tasks, schedule or workflow runs); sandboxes for jobs and worker pools in Preview Source for Cloud Run, Functions and one-off jobs
Security
IAM, audit log and 2FA in the free edition
Levelrail Yes. Allow/Deny IAM policies (stable, live e2e), passkeys and TOTP, and an audit log with a verifiable hash chain, CSV export and forwarding, all in the one edition Source for Levelrail, IAM, audit log and 2FA in the free edition
Cloud Run Partly. Cloud IAM and audit logs; 2FA is not a Cloud Run feature Source for Cloud Run, IAM, audit log and 2FA in the free edition
SAML or SCIM single sign-on
Levelrail Active. OIDC sign-in (17 named providers plus generic OpenID Connect) is on by default once configured. SAML and SCIM are beta, behind a toggle and off by default; SAML ran against Keycloak and SimpleSAMLphp in a local lab, not Okta or Entra Source for Levelrail, SAML or SCIM single sign-on
Cloud Run Not stated in their docs
Audit log of every action
Levelrail Yes. Every mutating request recorded; verifiable hash chain, CSV or NDJSON export, retention, streaming to a webhook or S3 bucket Source for Levelrail, Audit log of every action
Cloud Run Partly. Cloud Audit Logs cover Cloud Run API methods; the page does not state which are on by default, and Data Access logs appear to need enabling; pricing and retention not stated on the page Source for Cloud Run, Audit log of every action
Two-person deploy approvals
Levelrail Yes. A different person with the deploy ability must approve; the requester cannot approve their own. Approval also works from Slack or Discord Source for Levelrail, Two-person deploy approvals
Cloud Run Partly. Via Cloud Deploy, a separate product: requireApproval on a target, approver role roles/clouddeploy.approver. The page opened describes Cloud Deploy generally, not a Cloud Run page Source for Cloud Run, Two-person deploy approvals
As of 2026-11-18. Statements about other projects come from their own public documentation, linked on each cell, and can be out of date. Product names and logos belong to their owners and are used for identification only.
Questions people ask when they compare
What is the best self-hosted alternative to Coolify?
Levelrail. It is one binary that idles at about 200 MB and ships metrics, log search, IAM, audit, previews, rollback and a template catalog in a single free edition, with an importer that moves your Coolify apps and databases over.
What is a self-hosted alternative to Vercel, Heroku or Railway?
Levelrail gives you webhook-driven git deploys, preview environments, TLS (self-signed until you enable Let's Encrypt), logs, metrics and rollback on servers you own. It does not offer a global edge network, an edge runtime, ISR, image optimization, visitor analytics or a managed cloud. Functions exist but are container cold starts, so if you need the rest, a hosted platform is the better fit. Dokku is the closest minimal option for a single server.
Can I migrate from Coolify, Dokploy or CapRover to Levelrail?
Yes. Levelrail has a read-only importer with a dry-run mode, so you can keep the old platform running while you try the new one, then move DNS when you trust it. The migration guide in the docs walks through it.
How lightweight is Levelrail as a self-hosted deployment platform?
Levelrail is one Go binary for the control plane and one for the agent. On a 2 vCPU, 8 GB Linux test box the control plane idled at about 200 MB of memory and 1.5% of one CPU with no apps. Your apps use their own resources on top. Smaller figures (68 MB with no apps, 91 MB with 500 suspended apps) come from an Apple M4 Max development build and are not comparable. Both are single runs, and the method and limits are published in the docs.
Is Levelrail a good alternative to Coolify?
Yes. Levelrail does what Coolify does with a lighter footprint and more built in. An agent dials out over mTLS to the Docker Engine API instead of SSH from the control plane. A level-triggered reconciler records a status reason per pass, instead of Docker Compose per resource. Full-text log search and pinned-image rollback ship in the core. A read-only importer and a dry-run mode let you move without touching Coolify.
Is Levelrail a good alternative to Dokploy?
Yes. Levelrail does what Dokploy does with a lighter footprint and more built in. Plain Docker with a reconciler and a small agent per node, with no Swarm or overlay network. Full-text log search and crashloop diagnosis ship in the core with no extra stack. Dokploy now has metrics with threshold alerts. One Go binary with embedded SQLite and embedded Caddy, instead of Traefik. A read-only importer and a dry-run mode let you move without touching Dokploy.
Is Levelrail a good alternative to CapRover?
Yes. Levelrail does what CapRover does with a lighter footprint and more built in. Plain Docker containers with a level-triggered reconciler instead of Swarm services. An agent per node dials out over mTLS, so managed servers need no inbound port. Caddy embedded for ingress instead of nginx, with log search and crashloop diagnosis in the core. A read-only importer and a dry-run mode let you move without touching CapRover.
Is Levelrail a good alternative to Dokku?
Yes. Levelrail does what Dokku does with a lighter footprint and more built in. A dashboard with deploy history and one-click rollback to a pinned previous image in the free edition. Dokku Pro adds a paid web UI. Metrics and log search in the core, where Dokku ships log sinks and expects an external stack. More servers join with a one-time token and an agent that dials out. A read-only importer and a dry-run mode let you move without touching Dokku.
Is Levelrail a good alternative to Kamal?
Yes. Levelrail does what Kamal does with a lighter footprint and more built in. Desired state lives in a database, with a reconciler recording a status reason after every pass. Agents dial out, so CI uses a scoped API token instead of holding SSH keys. Adds a dashboard, deploy history, rollback and built-in observability, and is heavier by design. A read-only importer and a dry-run mode let you move without touching Kamal.
Is Levelrail a good alternative to Vercel?
Yes. Levelrail does what Vercel does with a lighter footprint and more built in. Runs on Linux servers you own, so secrets, metrics and logs stay on your nodes. Cost is the servers you rent, not per request or per seat. Container apps built from a Dockerfile. Vercel offers containers in beta too; Levelrail has no edge runtime or edge network, and functions sleep when idle. A read-only importer and a dry-run mode let you move without touching Vercel.
Is Levelrail a good alternative to Heroku?
Yes. Levelrail does what Heroku does with a lighter footprint and more built in. Runs on Linux boxes you own, with no per-dyno or per-app fee. Eight managed database engines with scheduled, verified backups. Heroku Postgres is first party with backups and rollback; other engines come as add-ons. Dockerfile builds with readiness-gated cutover, not buildpacks. A read-only importer and a dry-run mode let you move without touching Heroku.
Is Levelrail a good alternative to Railway?
Yes. Levelrail does what Railway does with a lighter footprint and more built in. Runs on your Linux servers, so pricing is your own server cost. Secrets, metrics and logs stay on your nodes unless you configure an export. Self-hosted software only; there is no managed cloud version. A read-only importer and a dry-run mode let you move without touching Railway.
Is Levelrail a good alternative to Cloud Run?
Yes. Levelrail does what Cloud Run does with a lighter footprint and more built in. A fixed-price server you run, so cost does not move with traffic. Autoscaling is in beta, and wake from sleep is slower than a cold start. Logs, metrics and data stay on machines you control, with nothing billed by volume. A read-only importer and a dry-run mode let you move without touching Cloud Run.
Run it next to what you have
One command installs the control plane. Apache 2.0, one edition, nothing gated.
docker pull ghcr.io/glincker/levelrail:latest