OpenClaw + MCP + Neo4j: Build an AI Agent for YouTube GraphRAG

In this tutorial, we bridge the gap between our Neo4j knowledge graph and our OpenClaw AI agent using a custom Model Context Protocol (MCP) server.

Building on our previous tutorials (where we used n8n to collect YouTube metadata and Gemini embeddings to store the graph in Neo4j), this video demonstrates how to make your AI agent "smarter" by giving it direct, standardized access to your graph database. We will walk through testing the MCP server locally, configuring environment variables, registering the server with OpenClaw, and creating a custom skill so your agent can answer questions natively using your own GraphRAG data.

🔗 Resources & Code:
All commands, .env templates, and complete code for this tutorial are available here:
👉 GitHub Repository: https://github.com/lbsocial/openclaw-n8n-neo4j-workflows
👉 LBSocial Official Website: https://www.lbsocial.net/post/openclaw-mcp-neo4j-youtube-graphrag-agent

⏱️ Timeline / Chapters:
00:00 - Introduction & Recap: n8n, Gemini Embeddings, and Neo4j
00:18 - Live Demo: Querying the YouTube GraphRAG Agent
01:02 - What is MCP? (Model Context Protocol Architecture)
02:59 - Setup: Cloning the MCP Repository
03:34 - Environment Configuration (uv sync and .env setup)
05:01 - Local Testing: Running the MCP Server tool menu
06:55 - Registering the MCP Server with OpenClaw
08:20 - Testing the Integration via Telegram
09:18 - Creating a Custom Skill for the AI Agent
10:06 - Final Test: Retrieving Video Recommendations
10:55 - Summary & Conclusion

💡 Don't forget to like and subscribe for more tutorials on AI agents, data science, and automation!

#OpenClaw #Neo4j #MCP #GraphRAG #n8n #AIAgents #DataScience #LBSocial #Gemini