Peen documentation

Peen is a coding agent built as a server. One controller opens a session for each workspace directory and runs the agent for that session in its own process or Docker container. Clients send work and watch it live over WebSocket, and the agent keeps working when they disconnect.

Every conversation, tool call, event, compaction, and provider exchange lands in SQLite, so you can reconnect after a restart and read what happened. After a crash, Peen restores saved context and reuses completed activity results.

The source is at github.com/psyb0t/peen.

Install

You need Linux, a model provider API key, and, for the default setup, Docker that your user can run.

curl -fsSL https://raw.githubusercontent.com/psyb0t/peen/main/install.sh | bash

Read install.sh before piping it into a shell. Getting started walks through the first run.

Find what you need

You want to	Read
Start from zero	Getting started
Install Peen, use the peen command, or run it by hand	Deployment
Decide where the agent runs and what it can touch	Where it runs
Know what the agent can reach before you deploy	Security
Configure providers, limits, and logs	Configuration
Run turns and manage sessions from a shell	Command line
Use the browser chat	Control surface
Run a prompt in a GitHub workflow	GitHub action
See what the agent can do	Agent tools
Teach the agent about your project	The harness
Build a client	WebSocket API and REST API
Embed Peen in a Go process	Embedding Peen
Let another agent operate Peen	Agent integrations
Fix something that broke	Troubleshooting
See how it fits together	Architecture
See how it does on coding benchmarks	Benchmarks
Change Peen itself	Developing Peen