All Aboard: Reasoning with LLMs and Neo4j via MCP
summary
Official MCP server connects Neo4j to LLMs so an agent runs queries, returns Cypher, and executes graph algorithms live. Demo uses a London Underground graph with 2772 stations, plans fastest routes (Baker Street to a work station), runs betweenness centrality via Neo4j GDS, and uses Claude Haiku 4.5 as the model. Server logs discovered tools, performs projections for GDS procedures, and even suggests schema changes to make graph queries more efficient.
AI-written summary · about
chapters (8)
- 0:00Introduction to MCP and LLMs
- 3:12Connecting LLMs to Neo4j via MCP
- 6:52Demo setup and London Underground dataset
- 8:20Graph summary and route planning
- 10:13Graph analysis with centrality
- 12:00Neighborhood clustering analysis
- 13:25Coffee shop location recommendation
- 14:32Improving graph schema with AI
related talks
53mAgents, Tools, and Mcp, Oh My! Next Level AI Concepts for Developers
46mMCP for LLM Agents, Infrastructure APIs & Data Services
28mGrounded Intelligence Boosting LLMs with Real-Time Graph Context
32mMulti-Agent Query Generation w/ LLMs on Complex Knowledge Graphs
45mAgents, Tools, and MCP, oh my! Next-level AI concepts for developers
28mTalk to your data without coding agentic AI and Claude with the MCP protocol
44mBeyond the Prompt: How MCP Connects Language Models to Real Tools
22mBuilding Cypher Queries With Open-Source LLMs