Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
- Added frontmatter with name, description, and tags to SKILL.md for improved metadata. - No changes to functionality or instructions—documentation only.
What This Skill Does
HTTP bridge that spawns MCP servers as persistent child processes and exposes their tools via REST endpoints. Designed for agents that lack native MCP support, where exec-based spawning kills servers when they go idle. Includes auto-restart on crash, rate limiting, IP allowlisting, and audit logging.
Solves the stdio persistence problem that kills MCP servers when spawned via exec, without requiring native MCP client support in the consuming agent.
When to use it
- Calling GitHub MCP tools from an agent that only supports HTTP
- Keeping MCP servers alive across long-running automation sessions
- Bridging multiple MCP servers under a single HTTP interface
- Auto-restarting a crashed MCP server during an overnight batch job
- Restricting MCP tool access to specific IPs on a shared machine
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: use the GitHub MCP server's search tool from an OpenClaw agent
- 1Register the server: `node cli.js add-server github npx @anthropic/mcp-github`
- 2Set credentials: `node cli.js set-env github GITHUB_TOKEN ghp_xxx`
- 3Start the bridge: `pm2 start bridge.js --name cherry-mcp`
- 4POST to `http://localhost:3456/github/call` with the tool name and arguments
- 5Receive the MCP tool response as JSON over HTTP
GitHub MCP tools are callable via HTTP REST from any agent or script
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.