clawd-mcp
FastMCP 2.14+ server bridging Cursor and Claude Desktop with the OpenClaw (openclaw.ai) and Moltbook (moltbook.com) ecosystem.
Overview
- OpenClaw: Personal AI assistant runtime; WhatsApp/Telegram/Discord gateway; Pi agent; bash, browser, cron
- Moltbook: Social network for AI agents; posts, comments, DMs, submolts; semantic search; heartbeat pattern
- clawd-mcp: MCP tools to invoke OpenClaw agents, manage sessions, skills, gateway; future Moltbook integration
Quick Start
pip install -e ".[dev]"
python -m clawd_mcp
cd webapp && npm install && npm run dev
Documentation
| Doc | Description |
|---|---|
| Index | Full documentation index |
| OpenClaw | OpenClaw platform, Gateway, channels, Pi agent, ClawHub |
| clawd-mcp Server & Tools | MCP server, portmanteau tools, dialogic returns |
| Webapp | React + Tailwind dark dashboard |
| Moltbook | Moltbook site, skills, agent heartbeats |
| Security Hardening | OpenClaw threats, hardening checklist, clawd_security |
Security
OpenClaw has major security risks (exposed creds, prompt injection, malicious skills). Use clawd_security for audit, skill scanning, config validation, and hardening recommendations. Optional CLAWD_MOUNT_VBOX=1 mounts virtualization-mcp for VM-based sandbox provisioning. See docs/SECURITY_HARDENING.md and Auth0, Intruder.
Configuration
| Variable | Default |
|---|---|
OPENCLAW_GATEWAY_URL | http://127.0.0.1:18789 |
OPENCLAW_GATEWAY_TOKEN | (required when auth enabled) |
MOLTBOOK_API_KEY | (optional) |
CLAWD_MOUNT_VBOX | 1 to mount virtualization-mcp at vbox_* (optional) |
Checks
.\scripts\check.ps1 -All # ruff, mypy, pytest
just check # same via just