Terminal-native AI peer
Kin documentation
A short path from installation to a first useful session, with clear links into Kin's complete maintained manual.
Available nowInstall
The supported installer works on macOS and Linux and can be run again safely.
bash <(curl -fsSL https://get.kinra.ai/install.sh)
kin doctorConnect a model
Run Kin, open the provider flow, and connect an OpenAI-compatible or Anthropic model route. The in-app flow keeps credentials out of project files.
- Start `kin` in a project directory.
- Open `/providers` and choose a provider.
- Verify the route before beginning work.
Begin with real work
A concrete task creates better calibration than a tour of features.
- Name the outcome and the constraints that matter.
- Let Kin inspect before prescribing a solution.
- Review the first shape and respond with specific judgment.
cd your-project
kinAuthority and verification
Kin exposes modes and permission gates so action stays proportional to risk.
- Use strict mode when you want each material action reviewed.
- Use auto mode for bounded local work with explicit escalation at protected boundaries.
- Read diffs and retain independent checks for consequential changes.
Maintained depth