clawpipe-mcp-server

Model Context Protocol server for ClawPipe — expose the intelligent AI pipeline as tools for Claude, ChatGPT, and other AI agents.

Tools

| Tool | Description | |------|-------------| | clawpipe_prompt | Send a prompt through the full pipeline (Booster, Pack, Cache, Route, Call, Learn) | | clawpipe_analyze_cost | Estimate cost of a prompt without sending it | | clawpipe_stats | Get current session telemetry snapshot | | clawpipe_booster_check | Check if Booster can resolve a prompt without AI |

Setup

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "clawpipe": {
      "command": "npx",
      "args": ["clawpipe-mcp-server"],
      "env": {
        "CLAWPIPE_API_KEY": "cp_xxx"
      }
    }
  }
}

Cursor

Add to Cursor settings → MCP:

{
  "mcpServers": {
    "clawpipe": {
      "command": "npx",
      "args": ["-y", "clawpipe-mcp-server"],
      "env": { "CLAWPIPE_API_KEY": "cp_xxx" }
    }
  }
}

Smithery

One-click install via smithery.ai/server/clawpipe — the smithery.yaml in this repo is the submission manifest.

Official MCP Registry

Submitted via server.json (root of this package). Search "clawpipe" in the MCP Registry.

Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | CLAWPIPE_API_KEY | Yes | Your ClawPipe API key | | CLAWPIPE_PROJECT_ID | No | Project ID (defaults to mcp-default) |

Publishing / registration

1. npm: npm publish --access public (package is clawpipe-mcp-server).
2. Official MCP Registry: submit server.json at https://registry.modelcontextprotocol.io/submit
3. Smithery: connect the GitHub repo at https://smithery.ai/new — smithery.yaml is auto-detected.
4. Glama + mcp.so: both index from GitHub topics. Ensure mcp topic is set on the repo.

Development

npm install
npm run build
npm test

License

MIT