Skip to content
atomic-agent docs

Installation

Requirements

atomic-agent runs on your own machine and drives real local tools. You’ll need:

  • Node.js (20+ recommended)
  • llama-server — a running local inference server, or let the CLI manage one
  • A Chromium-family browser — Chrome or Edge, for browser operation via playwright-core
  • git — for repository inspection tools

State (SQLite files and traces) is stored locally under a state directory on your machine.

Install

Terminal window
curl -fsSL https://api.atomicbot.ai/agent-install | sh

Provider modes

atomic-agent can talk to models in several ways:

  • Managed local llama.cpp — the default; the CLI runs a local server for you.
  • External llama-server endpoints — point at a server you already run.
  • OpenAI-compatible HTTP and OpenRouter-style providers — for text or embeddings, only when you deliberately configure them.
  • Tauri sidecar — newline-delimited JSON, for a native wrapper.
  • Telegram — single-user remote control with owner pairing and inline approval buttons.
  • MCP servers — connect external tools and expose them through the same registry.

See Configuration to set your provider and limits.