TypeClawDeploy
Run the agent on a server instead of your laptop — almost nothing changes
The reassuring part: deploying is the same thing you already do. There's no "production mode," no separate build, no special runtime. The typeclaw start you run on your laptop is the same command you run on a server.
The whole deployment
Copy the agent folder to your server and start it:
ssh server
cd /opt/my-agent
typeclaw startThat's it. Same folder, same config, same command. The container is the unit of deployment; the server is just a different place to launch it.
What actually changes between laptop and server is small and optional: where credentials come from, whether it restarts itself after a reboot, and whether you're running more than one agent. Each is a box below — open the ones you need.
Why deployment is boring (and that's the goal)
The host side of TypeClaw is just a launcher — it starts a container and talks to it. It holds no agent state and loads no plugin code. So there's nothing to "drift" on a server: it's a folder and a Docker image, the same on every machine. The reasoning is in How it's built.
Common server needs
That's the guide chain. From here, the Concepts explain the why behind the design, and the Reference has every command, flag, and config field.