Getting Started

JameClaw Docs

Install JameClaw, connect your model credentials, launch the local agent gateway, and run your first focused automation task.

JameClaw is an AI agent workspace for running useful local tasks. Start from source, configure one provider and one integration, then expand after the first successful run.

jameclaw install
jameclaw agent

Agents can

run focused local workflows from a browser-based launcher
connect to the model provider and credentials you configure
start a gateway for runtime tasks and tool access
summarize messages, inspect logs, and automate routine work

Choose your path

Use the local launcher for day-to-day work, or keep the static docs nearby when setting up a new environment.

Run from source

Clone the repository, install dependencies, and use the dev command while iterating locally.

Launch the agent

Open the UI, add credentials, configure the gateway, and give JameClaw one concrete task.

Install JameClaw

Download precompiled binaries directly from the official website [jameclaw.xyz](https://jameclaw.xyz/) or install using Node/npm package manager.

Get the binary

Download from the official portal or run npm install -g jameclaw.

Run onboarding

On fresh machines, run jameclaw install to set up the default workspace folders.

Configure settings

API keys, model provider fallbacks, and channel rules are managed under ~/.jameclaw.

Configure your first run

Open the browser Web UI at http://localhost:18800 by running jameclaw-webui or jameclaw-launcher. Visual controls allow you to add model credentials (OpenAI, OpenRouter, Anthropic, local model routes) and test connectivity.

Run your first task

Start the agent launcher session from your preferred entry point:

Web Console: Complete chat, provider routing, and logs interface at http://localhost:18800.
TUI Launcher: Run jameclaw-launcher-tui for headless servers or remote SSH sessions.
Terminal Agent: Run jameclaw agent for stateful command-line chat loops.

Gateway Configuration

The Gateway is the multi-channel background bridge connecting your agent to channels such as Telegram, Discord, Slack, Matrix, and LINE. Enable channels and manage credentials directly from the Web Console or TUI dashboard.

Provider Routing & Fallbacks

JameClaw allows you to manage multiple LLM provider accounts from a single configuration, routing requests and specifying fallbacks when primary endpoints go offline.

Supported providers include OpenAI, Anthropic, OpenRouter, Nous, Kimi/Moonshot, MiniMax, and local/custom OpenAI-compatible endpoints.

// Configuration schema in ~/.jameclaw/config.json
{
  "providers": {
    "openai": { "api_key": "sk-..." },
    "openrouter": { "api_key": "sk-or-..." }
  },
  "routing": {
    "default_model": "gpt-4o",
    "fallback_model": "openrouter/anthropic/claude-3.5-sonnet"
  }
}

Useful commands

jameclaw install          # Onboard fresh workspace config
jameclaw-launcher         # Start Go WebUI console launcher
jameclaw-launcher-tui     # Launch Headless SSH TUI launcher
jameclaw agent            # Interactive CLI agent session
jameclaw agent -m "..."   # Send a one-shot query message
jameclaw uninstall        # Wipe local runtime workspace state

Tips

Start with one model provider and one integration.
Make the first task narrow enough that success is obvious.
When debugging, check config first, then credentials, then gateway state.