Start here

Quickstart

Lampson is a coding agent that lives in your project. You open it inside a folder, tell it what you want in plain words, and it reads, edits and runs — showing you each step and asking before anything risky.

Install§

npm i -g lampson

No Node.js? There is a one-line installer for Windows, macOS and Linux that brings everything along.

Open a project§

cd path/to/your/project
lampson            # in the terminal
lampson --web      # in the browser, at http://127.0.0.1:8080

The folder you are in becomes the agent's workspace: the only place it can read or write. Both the terminal and the web version work on the same project, share the same settings and the same sessions.

Pick a model§

The first time, Lampson asks two things: which provider you use (DeepSeek, Anthropic, OpenAI, Kimi, Groq, OpenRouter, Ollama…) and your API key. That is saved on your machine — nowhere else — and shared by the terminal and the web UI. Change it whenever you like with /model.

Don't have a favourite? DeepSeek is inexpensive and works very well. Ollama runs a model on your own computer with no key at all. See Models and providers.

Ask for something§

Type what you want as you would to a colleague:

❯ the login form doesn't show an error when the password is wrong. fix it

You will see Lampson search, read the relevant files, edit one, and run the tests — each step on its own line with a ✓ and a short result. When it wants to do something with consequences (install a package, delete files, git reset) it stops and asks:

? bash npm install zod        allow? ❯ yes  no

Answer with the arrow keys, or p/d. Reading and editing files inside the project never need approval.

What next§

• Your first session — the terminal, / commands, ! to run things yourself.
• Permissions — what is allowed, what asks, what is always refused.
• Plugins — give the agent new tools and turn them on when you want.