Levelrail

Levelrail is an open source (Apache 2.0) self-hosted PaaS for developers and small teams who want a Vercel, Heroku or Railway style workflow on servers they own. It runs on any Linux server, from a small VPS to a rack (the one measured host is a 2 vCPU, 8 GB box), and turns plain servers into a private cloud with Git deploys (forge webhooks), HTTPS, databases with verified backups, metrics, logs and alerts, without Kubernetes and without the monitoring stack you would otherwise bolt on.

Status: pre-1.0 (v0.x). Single-server is production ready per Feature status; multi-node is beta.

Which install is for me?

The server installs on Linux only. macOS and Windows are CLI-only targets.

Related MCP server: Lizard MCP

Table of contents

Quickstart · How it works · Verify what you install · Why Levelrail · Production ready · Coolify alternative · Features · Guides · FAQ · For AI agents

Quickstart (install a self-hosted Vercel alternative)

You need a Linux server (amd64 or arm64) with systemd and root. The installer adds Docker if it is missing.

1. Install the server

curl -fsSL https://levelrail.com/install.sh | sudo sh

Note: HTTPS is self-signed until you enable Let's Encrypt. The first sign-in offers a one-click free certificate for <dashed-ip>.sslip.io, or add your own domain.

The last lines of output include the dashboard address and a one-time setup link:

Levelrail vX.Y.Z installed and running.

Open the dashboard:
  http://203.0.113.10:8080/login#setup=<token>

2. Create the admin account

Open the setup link and choose a password. The token sits in the # fragment, so it never reaches the server or its logs; lost it? sudo APP_DATA_DIR=/var/lib/levelrail-data levelrail setup-token prints it again. The first sign-in opens one screen that turns on a free certificate for <dashed-ip>.sslip.io with one click (no DNS needed) and deploys your first app.

3. Deploy an app

On the server, the installer already put levelrail-cli in /usr/local/bin. Sign in once, then import a repository (levelrail-cli apps create --interactive asks for everything step by step):

levelrail-cli login --api-url http://<server-ip>:8080 --code ABCD-EF23
levelrail-cli import https://github.com/your-org/your-app --deploy

Prefer containers, a pinned version or a source build? See Installing. Want it as light as Dokku, or everything on? install.sh --profile minimal|standard|suite, see Server profiles. Coming from another platform? See Migrate to Levelrail: Coolify, Dokploy, CapRover, Dokku and compose files each get a dry run first, and the old server keeps serving.

After setup, the dashboard home shows a Connect your CLI card. Press Show my command, paste the line into your terminal, and the CLI signs in with a single-use code that expires in minutes (no password, no approval step):

logged in as "you@example.com"
token tk_9fQ2x saved to ~/.config/levelrail-cli/credentials under profile "default", expires 2027-01-18

Signing in with your password from a new device needs approval: the CLI prints the dashboard page to approve it on and the number to type. login --device also still works.

Without --deploy, import only prints the deployment plan. To see the app's address, run levelrail-cli open your-app. It prints the custom domain first, then the automatic URL.

Or let your coding agent do it

Install the skill once, then ask your agent in plain words. It deploys, waits for the rollout to actually converge, reads the failure if there is one, and rolls back safely:

npx skills add https://github.com/glincker/levelrail --skill levelrail

Deploy this repo to my Levelrail and tell me when it is live.

The skill is also listed at skills.sh. Details: Agents.

If something fails

More fixes: Troubleshooting. Full install options (Docker image, pinned version, source build): Installing. Deploy from a local project with levelrail-cli up: One command quickstart. Server profiles (install.sh --profile minimal|standard|suite): Server profiles. Migrating from another platform? levelrail-cli import platform coolify|dokploy|caprover --dry-run shows what would be created first, and the old server keeps serving. Full guide: Migrate to Levelrail. Publishing an open-source project? Add a one-click button to its README: Deploy to Levelrail button.

Install the CLI only

For your laptop or a CI runner, to control a server that already runs Levelrail:

brew install glincker/tap/levelrail-cli
npm install -g levelrail-cli
curl -fsSL https://levelrail.com/install-cli.sh | sh
irm https://levelrail.com/install-cli.ps1 | iex

The first two work on any supported platform, the third is for macOS and Linux, the last for Windows PowerShell.

Run it on a free Arm VM

Levelrail's release binaries and images are built for arm64, so it runs on an Oracle Cloud Always Free Arm VM. You open two firewalls (the cloud network rules and the host firewall), then run the same installer. Oracle sets the free limits and its idle-instance rule, and both can change. Read the Oracle Cloud guide and Oracle's Always Free page first. App images vary on Arm, so check an image's platforms before installing a template.

How it works

flowchart LR
  C["Dashboard, CLI, MCP server"] --> API
  subgraph CP["Control plane (one Go binary)"]
    API["HTTP API"] --> R["Level-triggered reconciler"]
    R --> S[("SQLite state")]
    O["Built-in metrics and logs"]
    I["Embedded Caddy ingress and TLS"]
  end
  R -->|"gRPC, mTLS"| A["Node agents (dial out)"]
  A --> D["Docker Engine API"]
  I --> D

The control plane keeps desired state in embedded SQLite and a reconciler makes Docker match it, retrying safely because it works from observed state each time. Each node runs an agent that dials out to the control plane over mTLS and talks to the Docker Engine API directly, so nodes need no inbound port. Metrics and logs are stored per node and queried together. Caddy is embedded for ingress and certificates. Details: Architecture.

Verify what you install

Releases publish a Sigstore signature for checksums.txt and a build provenance attestation. The first line below resolves the latest release tag; set V to a tag yourself to verify an older release.

V=$(gh release view --repo glincker/levelrail --json tagName -q .tagName)
BASE=https://github.com/glincker/levelrail/releases/download/$V
curl -fsSLO $BASE/checksums.txt -O $BASE/checksums.txt.sigstore.json -O $BASE/levelrail-linux-amd64
cosign verify-blob --bundle checksums.txt.sigstore.json \
  --certificate-identity-regexp 'https://github.com/glincker/levelrail/\.github/workflows/release\.yml@.*' \
  --certificate-oidc-issuer https://token.actions.githubusercontent.com checksums.txt
sha256sum --check --ignore-missing checksums.txt
gh attestation verify levelrail-linux-amd64 --repo glincker/levelrail

The installer checks the checksums itself. Container images are on GHCR (ghcr.io/glincker/levelrail, multi-arch). More: Installing.

Why people pick Levelrail

• Lightweight. About 200 MB of RAM and roughly 1.5% of one CPU idle, measured on a 2 vCPU, 8 GB Linux test box with no apps under load and the default feature set (one run, 2026-10-19). Numbers and method.

• Agents, not SSH. Each node runs a small agent that dials out over mTLS and talks to Docker's Engine API directly. No inbound ports, no shelling out to the docker CLI, no polling loops.

• Observability included. Per-node metrics at 15 second resolution, searchable logs, deploy markers, traces, SLO burn, alert inbox and an idle-footprint health page. No Grafana stack to install.

• Deploys that protect you. Readiness-gated rollouts, pinned images for rollback, canary releases that roll back on errors, deploy diffs, approvals and freezes.

• Secure by default. Hash-chained audit log you can verify, SSRF-guarded outbound calls, envelope-encrypted secrets with key rotation, scoped tokens, passkeys and TOTP 2FA, OAuth and OIDC sign-in, IAM policies, signed releases. SAML single sign-on and SCIM provisioning are available as a beta feature, off by default.

• Built for AI agents. An MCP server with over 240 tools in on-demand toolsets (small default list, about 2.4k tokens; see tool surface), a scoped-token permission model and an agent skill.

Production ready, with receipts

Single-server deployments are production ready today. Multi-node and a handful of newer features are labelled beta, and every label links to its evidence.

Levelrail vs Coolify, Dokploy, CapRover and Dokku (Coolify alternative)

Most self-hosted platforms in this category manage servers by SSHing in and shelling out to the docker CLI, then parsing text, which tends to force polling loops. Levelrail uses a reverse-dialed agent and a level-triggered reconciler instead.

This is positioning, not a ranking, and every project here is worth using. Where Levelrail trails: it is the youngest project here, with fewer users and less production mileage, and its SAML single sign-on and SCIM provisioning are beta (details). Its template catalog (about 360 entries) is in the same range as Coolify's, but only a sample is deployed end to end (ADR 015). The full, sourced comparison is in Comparison.

What you get

• Deploys: Push to your git host (GitHub, GitLab, Bitbucket, Gitea webhooks), deploys start from a forge webhook (a native git push remote is beta and off by default), preview environments per pull request (stable on GitHub, beta on GitLab and Bitbucket), framework presets (Next.js, Go, React, Spring Boot and more), Dockerfile, Compose or image deploys, canary and blue-green rollouts. Readiness-gated cutover is live-tested with zero failed requests on a single node; multi-node is beta. Deploying apps

• Domains and TLS: Embedded Caddy ingress. The default certificate is internal (self-signed). Public Let's Encrypt is one toggle and was verified in one live run; renewal and DNS-01 wildcards are unverified. Custom certificates can be uploaded. Per-app abuse protection. Domains

• Databases: Postgres, MySQL, MariaDB, MongoDB, Redis, KeyDB, Dragonfly and ClickHouse with scheduled backups, Postgres point-in-time recovery, restore, and password rotation. Databases

• One search for everything: Cmd+K across apps, databases, nodes, templates and actions, filtered by what you may do. Search

• Servers: Add nodes with a one-time token, WireGuard mesh, zones and failure-domain spreading, placement what-if. Multi-node

• Templates: about 360 one-click services, with a smoke test over the catalog (101 of 108 tested passed in the 2026-10-25 and 2026-10-27 runs, status). Catalog

• Identity and access: Password, passkeys, TOTP, OAuth with 20 providers plus generic OIDC, IAM policies, teams, a hash-chained audit log with CSV export. SAML single sign-on and SCIM provisioning are beta and off by default. Identity and access, SAML and SCIM

• Plugins: A signed plugin marketplace with one-click install, dashboard panels and first-party alert channels. Plugins

• Automation: One HTTP API behind the dashboard, a CLI and an MCP server. AI never sits in the reconciliation path. CLI, AI control

• Pipelines, previews, cost budgets, sleep when idle, status pages and more. See Product tour.

More screens (metrics, logs, backups, domains, security, plugins) are in the Product tour.

All screenshots are captured from a real running control plane.

Guides

Step by step walkthroughs, each tested against a real install:

Add a Deploy button to your project

Maintain an open-source app? Let people run it on their own server in a few clicks.

Copy-paste snippets and the four badge styles: Deploy button and the Brand kit.

What is next

Planned, not promised. Dates are targets and items can move or drop. Full list in the roadmap.

• Real-host runs for the multi-node mesh and an idle-footprint number.

• Public ACME checks: renewal near expiry and DNS-01 wildcards.

• More upgrade and downgrade testing across releases.

• Release snapshots and environment bindings.

• A first stable release, only when the hardening evidence is in. No date is set.

FAQ

Is Levelrail production ready? Single-server deployments are production ready; multi-node and some newer features are beta. Every label is in Feature status. The project is pre-1.0.

Does it need Kubernetes? No. It is one Go binary with a Docker runtime, and a Kubernetes compatibility layer is a deliberate non-goal (Comparison).

What does it cost? Nothing. It is Apache 2.0, one edition, no license key and no paid tier. You pay for your own servers (Pricing).

How is it different from Coolify or Dokploy? Nodes are managed by a reverse-dialed agent and a level-triggered reconciler instead of SSH and the docker CLI, and metrics and logs are built in. See the comparison table, including where Levelrail trails.

Can I migrate from Coolify? Yes. levelrail-cli import platform coolify --dry-run shows what would be created first, and the old server keeps serving. Dokploy, CapRover, Dokku and compose files work too: Migrate to Levelrail.

Does it support AI agents and MCP? Yes. levelrail-mcp exposes the same API and permission model, and an agent skill is included. AI is never in the reconciliation path: MCP, AI control.

How do I upgrade? Use Settings > Updates or levelrail-cli upgrade --plan. The upgrade verifies the release, snapshots your data and rolls back by itself if the new release is not healthy: Upgrade safely.

Where is my data stored? On your servers. The control plane keeps SQLite, the master key and the setup token in /var/lib/levelrail-data by default; metrics and logs stay on each node (Installing, Architecture).

For AI agents and crawlers

• Agent skill: skills/levelrail/SKILL.md teaches a coding agent to deploy, wait, diagnose and roll back safely.

• MCP server: levelrail-mcp exposes the same API and permission model. Start in read-only mode. See MCP and AI control.

• Repo guide: AGENTS.md for working in this codebase, and llms.txt as the index. Docs for LLMs: llms.txt and llms-full.txt.

Community and support

• GitHub Discussions for questions, ideas and show and tell, and Issues for bugs.

• GLINR Discord: live chat with the maintainers, in the #levelrail forum channel.

• support@levelrail.com for direct help, and the Security policy to report a vulnerability privately.

• Contributing: dev loop, tests and commit conventions. Build from source needs Go 1.26 or newer and Docker: see Installing.

Star the repo

If Levelrail saves you a weekend, a star on GitHub helps other self-hosters find it. Bug reports and honest feedback help even more.

License

Apache 2.0, see LICENSE.

Built by

Levelrail is built by Gagan Deep Singh, founder of GLINR STUDIOS and theSVG. Sister projects: theSVG (brand icon library), theauth (open-source auth for AI agents and humans) and AskVerdict AI. Say hello at thegdsks.com.

Powered by GLINR STUDIOS · icon from theSVG