Skip to main content
Glama
miguelcc06

OpenClaw Docs MCP

by miguelcc06
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Hybrid

📖 OpenClaw Docs MCP

An MCP RAG server providing powerful hybrid search over OpenClaw documentation

OpenClaw Docs MCP indexes the OpenClaw documentation into PostgreSQL using pgvector and exposes powerful hybrid search tools via HTTP using FastMCP.

✨ Key Features

  • 🧠 Hybrid Search: Combines semantic (pgvector) and full-text search across ~721 English documentation pages.

  • 🛠️ 13 MCP Tools: Fully equipped with search, get_chunk, navigation (next/prev), page outlines, and index statistics.

  • ⚡ HTTP Transport: Exposed securely over HTTP with API key authentication.

  • 🔄 Incremental Sync: Smart content-hash syncing to avoid redundant updates.

  • ⚙️ Production Ready: Includes systemd service files and an optional daily sync timer.

Related MCP server: knowledge-server

📋 Requirements

  • Python: 3.11+

  • Database: PostgreSQL 15+ with the pgvector extension.

  • AI: OpenAI API key (for embedding generation using text-embedding-3-small).

🚀 Setup & Installation

1. Environment Setup

python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

2. Configuration

cp config.env.example config.env
# Edit config.env with: DATABASE_URL, OPENAI_API_KEY, API_KEY

3. Database & Ingestion

# Run migrations (ensure your database exists)
alembic upgrade head

# Crawl sitemap and index content (~5-15 min)
openclaw-docs-mcp ingest --full

4. Run the Server

openclaw-docs-mcp serve

📍 Server runs at http://0.0.0.0:8000/mcp (configurable via MCP_PORT).

🧰 MCP Tools Available

⚠️ Note: Write all queries in English for the best results (docs are indexed in English only).

Tool

Description

search

Hybrid semantic + full-text search

search_in_page

Search within a single document

lexical_search

Full-text search only

get_chunk

Full chunk by ID

get_chunks

Batch retrieve chunks

get_chunk_next / get_chunk_prev

Navigate document order

get_chunk_neighbors

Previous + current + next

get_context_window

Symmetric context around a chunk

get_document_chunks

Paginate chunks in a document

list_pages

List indexed pages

get_page_outline

Headings + chunk IDs

get_stats

Index statistics

🔌 Cursor Configuration

🏠 Localhost (same machine)

{
  "mcpServers": {
    "openclaw-docs": {
      "url": "http://127.0.0.1:8000/mcp",
      "headers": {
        "Authorization": "***"
      }
    }
  }
}

🌍 Remote (e.g., Cloudflare tunnel)

{
  "mcpServers": {
    "openclaw-docs": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote@latest",
        "https://your-tunnel.example.com/mcp",
        "--header", "Authorization: Bearer ${OPENCLAW_DOCS_API_KEY}"
      ],
      "env": {
        "OPENCLAW_DOCS_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

⚙️ systemd Deployment

For automated management and background syncing:

sudo cp deploy/openclaw-docs-mcp.service /etc/systemd/system/
sudo cp deploy/openclaw-docs-mcp-sync.service deploy/openclaw-docs-mcp-sync.timer /etc/systemd/system/
sudo cp deploy/sync-wrapper.sh /home/miguelcc06/openclaw-docs-mcp/deploy/
chmod +x deploy/sync-wrapper.sh

sudo systemctl daemon-reload
sudo systemctl enable --now openclaw-docs-mcp
sudo systemctl enable --now openclaw-docs-mcp-sync.timer

💡 Tip: Set SYNC_ENABLED=true and SYNC_SCHEDULE=daily in config.env.

💻 CLI Commands Quick Reference

openclaw-docs-mcp serve          # Start HTTP MCP server
openclaw-docs-mcp ingest --full  # Full re-index
openclaw-docs-mcp sync           # Incremental sync
openclaw-docs-mcp stats          # Index statistics
alembic upgrade head             # Apply DB migrations

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityStale
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers