# Lampson

> A free, open-source AI coding agent that lives in your project, written in Synsema. Terminal + web UI, tools confined to the mounted project, any OpenAI- or Anthropic-compatible model.

Install: `npm i -g lampson` · repository: https://github.com/kitecosmic/lampson · site: https://lampson.org/en

## Documentation

- [Quickstart](https://lampson.org/en/docs/quickstart.md): Install Lampson, open a project, pick a model, and give it your first task — in about two minutes.
- [Your first session](https://lampson.org/en/docs/first-run.md): How a Lampson session reads — the steps, the diffs, the commands that start with a slash, and running things yourself with an exclamation mark.
- [The web UI](https://lampson.org/en/docs/web-ui.md): lampson --web — the same agent in your browser, with a file tree, live process logs, real terminals, and panels for everything you can also do from the command line.
- [Permissions](https://lampson.org/en/docs/permissions.md): What Lampson does on its own, what it asks you about, and what it refuses no matter what — plus the three modes: ask, yolo and strict.
- [Models and providers](https://lampson.org/en/docs/providers.md): Which model APIs Lampson works with, how to switch, and how to point it at any compatible endpoint including Ollama on your own machine.
- [Workspaces](https://lampson.org/en/docs/workspaces.md): Every project you open is a workspace with its own process; a small hub at 127.0.0.1:8080 lets you switch between them and keeps them alive.
- [Sessions and traces](https://lampson.org/en/docs/sessions.md): Every conversation is saved with every step; reopen one later, and read the trace to see exactly what the agent did and how long it took.
- [Project memory](https://lampson.org/en/docs/memory.md): Lampson keeps notes per project — how to run it, gotchas, decisions — and reads them back next session. You can read and edit them.
- [Plugins](https://lampson.org/en/docs/plugins.md): Plugins are small folders that give Lampson new tools — a database query, your company's API, a deploy — in any language. Off by default; you turn each one on.
- [Skills](https://lampson.org/en/docs/skills.md): how we cut a release of this project
- [MCP servers](https://lampson.org/en/docs/mcp.md): Connect MCP servers with the same JSON you use elsewhere; their tools join the agent's toolbox and ask before running.
- [Language servers](https://lampson.org/en/docs/lsp.md): Give the agent go-to-definition, references and file outlines through your project's language servers, instead of grep guessing.
- [Sub-agents](https://lampson.org/en/docs/subagents.md): The agent can split work across parallel workers, or send a job to the background and keep talking to you; each worker has its own live log.
- [Scheduled tasks](https://lampson.org/en/docs/schedules.md): Run things with nobody at the keyboard — every 6 hours, daily at 9, weekdays at 8:30 — and approve anything risky from your phone.
- [Commands and flags](https://lampson.org/en/docs/commands.md): Every slash command of the terminal, every launch flag, and the keyboard shortcuts of the prompt.
- [Configuration](https://lampson.org/en/docs/configuration.md): Everything is optional. Where settings live, what .env can override, and the few knobs worth knowing about.
- [Docker and servers](https://lampson.org/en/docs/deploy.md): Run Lampson in a container, or on a VPS behind your own domain, with approvals from your phone.
- [FAQ](https://lampson.org/en/docs/faq.md): Short answers to the questions people ask first.