WeClaw (wemcp) is an MCP server that hosts and monetizes MCP tools and AI skills. It provides a unified platform to browse, import, debug, and run MCP servers and skills in real-time, with a hosted proxy layer that exposes local stdio servers via remote HTTP endpoints. Use the built-in admin panel to manage servers, set visibility, and classify them into categories, and leverage the agent chat to interact with your tools directly in the browser. To get started, deploy WeClaw via Docker as described in the Quick Start, then access the web UI to import MCP server configurations or npm/pip packages, and use the hosted proxy to test tools without client-side runtime.

Prerequisites:

• Docker 19.03.9+ and Docker Compose 1.25.1+ per WeClaw prerequisites
• Git
• Internet access to pull images

Installation steps:

1. Clone the repository: git clone https://github.com/tomeai/weclaw.git cd weclaw

2. Set up environment variables and config (example): cp backend/deploy/.env.server.example backend/deploy/.env.server

   edit the .env.server file as needed

3. Start services with Docker Compose (from the repository root): cd docker docker compose up -d

4. Access the UI:

   • Open http://localhost:8000 (or the configured host/port)
   • Sign in with GitHub OAuth as configured in the admin panel

5. Optional: import and run an MCP server or npm/pip package from the UI or by providing a config snippet. The Quick Start guide above demonstrates cloning and starting the environment.

Tips and common issues:

• Ensure Docker and Docker Compose versions meet the minimum requirements in the Prerequisites section.
• If the proxy cannot reach local STDIO servers, check your network configuration and that the hosted proxy (Caddy 2) is properly configured in docker-compose.
• Review the .env.server file for environment-specific settings such as database connections and Redis endpoints.
• When importing servers, you can use stdio/SSE/Streamable-HTTP configurations. The system will auto-compile metadata and present it in the UI.
• For troubleshooting, check container logs with docker logs <container_id>, and ensure that the MCP runtime versions (fastmcp, mcp SDK) are compatible with the server you deploy.