DanClaw

Multi-agent communication and automation platform. Routes messages from multiple channels (terminal, Slack, phone) through a central dispatcher to AI agents with configurable personas, tools, and permissions.

Architecture

• Listeners (listeners/): Channel adapters that translate external messages to a standard internal format
• Dispatcher (dispatcher/): Core routing, session management, permission resolution, and AI execution
• Tools (tools/): Standalone scripts invokable by agents (e.g., Obsidian vault access, git operations)
• Config (config/): JSON configuration for agents, permissions, and listeners; config loading and validation
• Personas (personas/): Markdown system prompts defining agent behavior

Tech Stack

• Python 3.11+, asyncio
• SQLite via aiosqlite
• aiohttp for HTTP-based IPC
• Docker Compose for deployment
• Tailscale for networking

Getting Started

# Copy environment template
cp .env.example .env
# Edit .env with your actual secrets

# Run with Docker Compose
docker-compose up

Status

Phase 1: Project scaffold in progress.