Summary

HappyClaw is a self-hosted, multi-user AI Agent system built on the Claude Agent SDK, directly leveraging the Claude Code CLI runtime. It offers unified access across Feishu, Telegram, QQ, and Web, providing robust multi-user isolation and automatically inheriting Claude Code updates. Ideal for power users and teams needing flexible, secure, and extensible AI agent infrastructure accessible from any device.

How It Works

The system directly utilizes the Claude Code CLI runtime, ensuring automatic adoption of Claude Code upgrades without adaptation. It features per-user isolation for workspaces, IM channels, and RBAC, alongside invite-code registration and audit logs. Messages from various channels are routed to a main process managing a concurrent queue for host (admin) or Docker container (member) agent execution. Agents communicate via IPC for 12 MCP tools, enabling features like messaging, scheduling, and memory management.

Quick Start & Requirements

• Prerequisites: Node.js (>= 20), Docker (for container mode), Claude API key.
• Installation: Clone the repo, then run make start for a one-click setup.
• Configuration: Guided Web UI setup for admin credentials, Claude API, and optional IM channels.
• Documentation: README serves as the primary guide.

Highlighted Details

• Multi-Channel Integration: Feishu, Telegram, QQ, and a responsive Web interface with mobile PWA.
• Agent Execution: Host (admin) or container (member) modes, multi-session concurrency, auto-recovery.
• Real-time Experience: Streaming updates for agent thoughts, tool calls, execution status, live Markdown, and web terminal.
• Extensibility: 12 MCP tools, comprehensive memory system, flexible skills.
• Security: User isolation, RBAC, encrypted API keys, mount security, login protection.

Maintenance & Community

Information on contributors, sponsorships, community channels (Discord/Slack), or a public roadmap is not detailed in the provided README.

Licensing & Compatibility

• License: MIT.
• Compatibility: Permissive MIT license allows commercial use and integration into closed-source projects.

Limitations & Caveats

Multi-channel integration requires third-party API credentials. Member users require Docker and a built container image for isolated execution, while admin users operate in host mode without Docker.