Self-hosted AI agents, on a server you own
Casama runs on your own hardware with one command. Your chats, memory and files stay on your machine, you choose the models, and it costs nothing.

What you get
A self-hosted Casama is the full product, not a cut-down edition:
- Bots with their own computers. Each bot works on a sandboxed Linux machine with a shell, files and a browser. You can watch it live and take over.
- Any model. Bring keys for Anthropic, OpenAI, Google Gemini, xAI or OpenRouter, sign in with your ChatGPT plan, or point a bot at a local model in Ollama, LM Studio or vLLM. Local models are recorded at $0.
- Approvals. Risky steps wait for your OK. You set the rules per bot and per action.
- Routines. Bots start on a schedule, on a webhook, on an email or a calendar event.
- Apps everywhere. The same apps for iPhone, iPad, Android, Mac, Windows, Linux and the web sign in to your server.
- No limits. As many users, bots and computers as your hardware can carry.
Install it
You need a machine with Docker. To try it on your own computer:
git clone https://github.com/stackwares/casama && cd casama
./install.sh --local
For a server other people reach over the internet, point a domain at the machine first:
./install.sh --domain casama.example.com
The installer checks Docker, creates every secret, starts the stack and waits until it is healthy. With Docker already installed it takes about 15 minutes, most of it building images. The quick start walks through signing in and your first bot.
Prefer no terminal? The desktop app for Mac and Linux can run a server on the computer it is installed on: choose Run Casama on This Computer on the sign-in screen.
Use it from your phone
A server on a laptop or a home network normally can't be reached from outside. Casama Link fixes that for free: your server gets its own casama.link address, with no domain, port forwarding or VPN to set up. Scan a code with the Casama app to sign in, and notifications reach your phone like any other app's. Your chats, memory and files stay on your server.
Self-hosted or Casama Cloud?
| Self-hosted | Casama Cloud | |
|---|---|---|
| Price | Free | Free, then $15, $49 or $99 a month |
| Where your data lives | Your machine | Casama's servers |
| Models | Any provider, your ChatGPT plan, or local models | Casama picks and runs them |
| Model costs | You pay providers directly; local models are free | Paid from credits |
| Setup | One command and Docker | Sign in |
| Upkeep | You update and back up | Casama does |
You can also mix them: keep your server and let Casama Cloud run the bot computers and model calls for it.
Self-hosting is free and stays free. If Casama earns a place on your server, you can sponsor the project on GitHub.
Questions
What hardware does it need?
One machine with Docker runs everything: the database, the API, the worker and the bots' computers. The docs list the requirements and production advice.
Is anything held back from the open-source edition?
The app is open source under AGPL-3.0. A separate ee/ folder holds a few enterprise features under their own licence; Casama builds and works without it.
Can I run it with only local models?
Yes. Point a bot at an OpenAI-style endpoint such as Ollama. Nothing has to leave your network.