claude code
anthropic’s coding agent does the work in each colony, and asks you with clear choices when it needs you.
- status
- built-in ships with colonizer and is on without you doing anything.
- version
- claude agent sdk 0.3.270
- licence
- mit (part of the harness); claude code is anthropic’s
- docs
- module readme ↗
what it does for you
the default settler. it runs claude code through the claude agent sdk inside the colony’s microvm. when it needs a decision it asks a multiple-choice question in the cockpit instead of writing an essay.
it can also run on another provider: point its models at deepseek, a server on your own network, or any anthropic-compatible endpoint. it is the only settler that loads skill packs and token savers.
what it needs
- a claude subscription (sign in once with claude setup-token) or an anthropic api key, kept on your machine
- on linux, native claude code installed on the machine; colonies run it read-only
how to turn it on
nothing to do: it is the agent every colony uses unless you pick another in settings → modules.
how far it has been tested
the only settler that ci runs in a real colony end to end; colonies built with it have opened real pull requests.