Kamal

Open source alternative to Fly.io, Heroku, Render, Railway, DigitalOcean App Platform and Porter

14.7K stars
MIT
Active this week
Last commit 1 day ago

Last verified Oct 22, 2026

Deploy containerized web apps to bare metal or cloud VMs with zero downtime using Docker and SSH, without a managed platform or vendor lock-in.

Visit website GitHub14.7K Deploy on Hostinger

Read the full breakdownWho it is for, how it works, trade-offs, the quick start and the FAQ

who it's for

Who Kamal is for#

Teams migrating off Heroku or Render to cut costs

Kamal gives you Heroku-like deployment commands (push to deploy, rolling restarts, zero downtime) on servers you rent at commodity prices. A team moving from Heroku's Performance dynos to a Hetzner or DigitalOcean server can cut hosting costs significantly while keeping the same deployment workflow.

Skip if:

If your team has no Linux or Docker experience and no interest in managing servers, a managed platform is still the better choice. Kamal reduces deployment complexity but does not eliminate server management.

Rails developers deploying to their own infrastructure

Kamal was built by the Rails team and is used in production at 37signals to run HEY email and Basecamp. It supports the standard Rails asset pipeline, web and worker process splits, and database migration hooks, making it a first-class deploy tool for Rails apps on servers you own.

Skip if:

If your Rails app is small enough that a low-cost managed instance covers your needs, the overhead of managing your own server is probably not worth it yet.

DevOps teams building cloud-agnostic deploy pipelines

Kamal's config file takes a list of IP addresses, making it trivially portable across cloud providers. Teams that want to avoid committing to a single cloud vendor, or that want to burst to a second cloud during traffic spikes, can use Kamal to maintain a provider-agnostic deploy pipeline.

Skip if:

If your deployment requirements include auto-scaling, complex load balancing, or multi-region failover at the infrastructure level, a Kubernetes-based setup or managed Kubernetes service is a better fit.

the problem

The problem it solves#

Managed deployment platforms like Heroku, Fly.io, and Render make it easy to ship web apps, but the cost of that convenience compounds over time. As traffic grows, monthly hosting fees on managed platforms can reach several times what equivalent server capacity would cost from a bare-metal or low-cost cloud provider. Teams that want to migrate off those platforms face a real problem: the tooling they relied on for deployments is tied to the platform, and rebuilding the deployment pipeline for self-managed servers has historically required learning Kubernetes or Docker Swarm.

Neither option is ideal for most teams. Kubernetes is powerful but operationally heavy, demanding dedicated infrastructure expertise to run safely. Docker Swarm is simpler but uses a declarative model that is harder to debug than direct imperative commands. Teams that just want to deploy a containerized app to a cheap server are caught between expensive managed platforms and complex self-hosted orchestration tools.

how Kamal solves it

How it solves it#

Zero-downtime rolling deploys

Kamal deploys a new Docker image alongside the running container, waits for the health check to pass, then switches traffic atomically. Requests never hit a container that is not yet ready. Rolling restarts replace containers one server at a time, so multi-server setups stay live throughout the deploy.

SSH-based server provisioning

Point Kamal at any vanilla Ubuntu server with an SSH key installed and it auto-provisions Docker, pulls your image, and starts the app. No Ansible playbooks, no pre-baked images, and no cloud-specific tooling. A server can go from fresh to running in minutes with no manual preparation.

kamal-proxy for traffic switching

Kamal includes kamal-proxy, a purpose-built proxy that handles traffic switching between old and new containers during deployments. It serves as the HTTP entry point, managing health checks and routing so your app stays available during every deploy.

Accessory service management

Kamal can manage accessory services (databases, caches, queues) on your servers alongside your main app container, using the same Docker-based model. You define accessories in your config file and deploy them with the same commands as your app.

Remote Docker builds

Kamal supports building Docker images on a remote builder machine rather than locally, which speeds up CI and deployment pipelines for teams with slow local machines or when building images for different server architectures.

strengths · trade-offs

Strengths and trade-offs#

Strengths

MIT license with full provider portabilityKamal is MIT-licensed and has no dependency on any hosting provider. You run it against servers from any provider, including Hetzner, OVH, DigitalOcean, or your own bare metal. Switching providers means updating a list of IP addresses in the config file, not rebuilding your deployment pipeline.
No server prerequisites beyond SSH accessKamal requires nothing more than a fresh Ubuntu server with an SSH key. Docker is installed automatically. No pre-built machine images, no Ansible playbooks, and no cloud-provider-specific tooling. This makes it practical to provision new capacity quickly across multiple clouds.
Production-validated by 37signals at scaleKamal was built by 37signals to run HEY email and Basecamp on their own hardware after migrating off cloud platforms. It is not a hobby project: it is the deploy tool 37signals uses in production every day, which means its sharp edges have been worked out against real high-availability workloads.
Simpler mental model than KubernetesKamal uses imperative commands: each deploy step is a direct action you can trace, not a diff against a desired cluster state. Teams familiar with Capistrano or Heroku's CLI find the model familiar. For the typical web app running on a few servers, Kamal's complexity ceiling is far lower than Kubernetes.

Trade-offs

-Requires Ruby to run deploysKamal is written in Ruby and must be installed as a Ruby gem. Teams without Ruby in their toolchain need to add it to their local environment or CI setup. The servers themselves do not need Ruby, but the machine that runs Kamal deploys does. A Docker-based runner is available for CI environments that prefer not to install Ruby directly.
-Not designed for complex orchestration at scaleKamal's imperative model works well for single-app and small multi-server deployments. It does not provide Kubernetes features like auto-scaling, complex service mesh configurations, or cluster-level resource management. Teams running hundreds of containers across dynamic node pools need a container orchestrator, not Kamal.
-Young project with 162 open issuesWith 162 open issues at time of writing, there are real rough edges to be aware of. Many stem from edge cases in Docker networking, proxy configuration, and multi-region deployments. The project is actively maintained, but it is young and may require workarounds on less common infrastructure setups.

versus alternatives

Kamal vs alternatives#

Kamal vs Heroku

Heroku and Kamal both simplify web app deployment, but they operate on opposite ends of the control spectrum. Heroku is a fully managed platform that hosts your containers; Kamal is a deployment tool you run against servers you choose.

Feature	Kamal	Heroku
License	MIT	Proprietary
Self-hosting	Yes	No
Server control	Full	None
Cost model	Server costs only	Per-dyno monthly fee
Zero-downtime deploys	Yes	Yes (with paid dynos)

Kamal wins on cost for teams running at non-trivial scale. Heroku Performance dynos are priced at a significant premium; comparable Hetzner or DigitalOcean servers cost a fraction of that, and you can run multiple apps on each. The tradeoff is operational overhead: Heroku handles database backups, SSL certificates, log drains, and scaling automatically, while Kamal leaves those responsibilities to you.

Kamal vs Fly.io

Fly.io is the more modern managed platform comparison. Both offer zero-downtime deploys and multi-region support, but Fly.io is a managed service and Kamal is not.

Feature	Kamal	Fly.io
License	MIT	Proprietary
Self-hosting	Yes	No
Multi-region	Manual (multiple IP addresses)	Built-in
Pricing	Server costs	Per-machine billing

Fly.io is a strong choice when you need multi-region deployment with minimal configuration and want to stay on a managed platform. Kamal is the better choice when cost control and the ability to run on your own hardware are higher priorities. Teams that have outgrown Fly.io's pricing model often find Kamal's approach more economical at scale.

install · quick start

Quick start#

bash

Install Kamal as a Ruby gem, then run setup to provision your servers and start the first deploy.
```bash
gem install kamal
kamal setup
```

tech stack · detected from GitHub

What it's built on#

Languages: Ruby

frequently asked

FAQ#

Is Kamal free to use?

Yes. Kamal is MIT-licensed and free to use with no restrictions. You install it as a Ruby gem and run it against servers you rent or own. The cost is your server infrastructure, not any Kamal licensing fee. There is no paid tier or managed offering from the Kamal team.

What operating system do my servers need to run Kamal?

Kamal officially targets Ubuntu and other Debian-based Linux servers. You need a fresh server with SSH key access; Kamal installs Docker automatically on first setup. Other Linux distributions work in practice but are less tested. Windows servers are not supported.

Does Kamal work with non-Rails apps?

Yes. Kamal deploys any app that can be packaged as a Docker image. It was originally built for Rails but works with Node.js, Python, Go, and any other web framework. The only requirement is that your app runs in a container and exposes an HTTP endpoint for health checks.

How does Kamal compare to Kubernetes for web app deployment?

Kamal is significantly simpler than Kubernetes for typical web app deployments. Kubernetes provides features like auto-scaling, service mesh, and complex scheduling that most web apps do not need. Kamal uses direct SSH commands and Docker, which is easier to understand and debug. The tradeoff is that Kamal does not scale to the cluster-management complexity that Kubernetes handles.

Can I use Kamal with any cloud provider?

Yes. Kamal works with any server accessible over SSH, regardless of provider. Common choices include Hetzner, DigitalOcean, OVH, AWS EC2, and bare metal colocated servers. You provide a list of IP addresses in the config file and Kamal handles the rest. Switching providers means updating that IP list.

also worth a look

Similar open-source tools#

CapRover

Alternative toHeroku

Self-hosted PaaS: deploys any language, automatic HTTPS, zero fees.

15,187 GitHub stars
Licence Apache-2.0 with additional terms

Dokploy

Alternative toHeroku

Self-hosted PaaS to deploy apps and databases across multiple servers

37,935 GitHub stars

openship

Alternative toHeroku

Self-hosted deployment platform with built-in CI/CD

14,865 GitHub stars
Licence Apache-2.0

Void Cloud

Alternative toVercel

Full-stack Vite toolkit with one-command Cloudflare deployment

507 GitHub stars
Licence MIT

Keda

Alternative toKedify

Event-driven Kubernetes autoscaling with 70+ scalers and scale to zero

10,646 GitHub stars
Licence Apache-2.0

Herdr

Alternative toConductor

Agent runtime that keeps terminals alive across machines

44,852 GitHub stars
Licence Apache-2.0

by category

Run Kamal on your own server

Self-host Kamal on a Hostinger VPS.

Deploy on Hostinger

Featured

rakazo

Alternative toGrok Bot

AI teammates you own: your keys, your model, your machine.

4,904 GitHub stars
Licence Apache-2.0
#AI Agents

Omnara

Alternative toClaude Managed Agents

Open-source agent deployment API. Self-host or use Omnara Cloud.

2,957 GitHub stars
Licence Apache-2.0
#AI Agents

Inbox Zero

Alternative toSuperhuman

AI email assistant that drafts replies and clears your inbox

12,712 GitHub stars
Licence AGPL-3.0
#Email

Feature your project

The projects we add and every new guide. Free, unsubscribe any time.

Find open source alternatives to paid software

Kamal vs Heroku

Feature	Kamal	Heroku
License	MIT	Proprietary
Self-hosting	Yes	No
Server control	Full	None
Cost model	Server costs only	Per-dyno monthly fee
Zero-downtime deploys	Yes	Yes (with paid dynos)

Kamal vs Fly.io

Fly.io is the more modern managed platform comparison. Both offer zero-downtime deploys and multi-region support, but Fly.io is a managed service and Kamal is not.

Feature	Kamal	Fly.io
License	MIT	Proprietary
Self-hosting	Yes	No
Multi-region	Manual (multiple IP addresses)	Built-in
Pricing	Server costs	Per-machine billing

Who Kamal is for#

Teams migrating off Heroku or Render to cut costs

Skip if:

Rails developers deploying to their own infrastructure

Skip if:

DevOps teams building cloud-agnostic deploy pipelines

Skip if:

The problem it solves#

How it solves it#

Zero-downtime rolling deploys

SSH-based server provisioning

kamal-proxy for traffic switching

Accessory service management

Remote Docker builds

Strengths and trade-offs#

Strengths

Trade-offs

Kamal vs alternatives#

Kamal vs Heroku

Kamal vs Fly.io

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

DevOps & CI/CD

Cloud & Hosting

Developer Tools

Find open source alternatives to paid software

Who Kamal is for#

Teams migrating off Heroku or Render to cut costs

Skip if:

Rails developers deploying to their own infrastructure

Skip if:

DevOps teams building cloud-agnostic deploy pipelines

Skip if:

The problem it solves#

How it solves it#

Zero-downtime rolling deploys

SSH-based server provisioning

kamal-proxy for traffic switching

Accessory service management

Remote Docker builds

Strengths and trade-offs#

Strengths

Trade-offs

Kamal vs alternatives#

Kamal vs Heroku

Kamal vs Fly.io

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

DevOps & CI/CD

Cloud & Hosting

Developer Tools

Find open source alternatives to paid software