Getting started

Install, key setup, first run, quickstart path.

Fastest path from zero to chatting with an agent that can read, edit, and run your code.

Prerequisites

Node.js >=18 (see engines in package.json)
A provider key. Default provider is OpenCode Zen: get one at https://opencode.ai/auth
A TTY for npm start (the TUI needs an interactive terminal)

Install

Global install gives you the atom binary:

npm i -g atom-agent
atom

Or run from source:

npm install

First run

Get a key at https://opencode.ai/auth
Set it for the session (PowerShell shown; use export on POSIX):

$env:OPENCODE_ZEN_API_KEY="sk-your-key"
npm start

Type / to see every command. Type /provider to paste a key once and store it instead of using env vars.

No key yet: the TUI still starts. Chatting without a key for the active provider errors inline and points at /provider. Nothing is posted.

Quickstart path

prerequisites
-> npm install
-> set OPENCODE_ZEN_API_KEY (or paste via /provider)
-> npm start
-> type / to list commands, ask something about your repo

Expected result: streaming answer with live tool activity and a status line showing provider, model, token usage, reasoning effort, and mode.

Next steps

CLI and TUI for slash commands and keyboard control
Providers and Models to switch off Zen or use a local OpenAI-compatible server
Configuration for all env knobs and the auth file
Troubleshooting if the first run fails

CLI and TUI →