UFO MCP
in progressKnowledge graph over UAP/UFO data — Neo4j + Qdrant, queryable via MCP. Because why not.
Neo4j · Qdrant · Python · MCP
A knowledge graph over 70 years of UFO/UAP reports from NUFORC, with an MCP interface so you can query it from any AI assistant that supports MCP.
Why this exists
The NUFORC dataset is genuinely interesting if you approach it as a data engineering problem rather than a belief question. ~150,000 reports, spanning 1974–present, with location, description, shape, duration, and witness count.
A SQL database answers simple questions. A knowledge graph + vector store answers the interesting ones.
Architecture
Neo4j stores the graph: reports as nodes, with relationships to locations, shapes, dates, and entities extracted from descriptions (aircraft types, colors, behaviors).
Qdrant stores embeddings of the free-text descriptions, enabling semantic search across reports without exact keyword matching.
MCP tools expose the graph to AI assistants. The tools are designed to be composable — you can chain them to answer complex queries.
Current state
Data ingested and cleaned. Graph populated. MCP tools working locally. Working on cloud deployment.