Senzing MCP Server

Senzing does entity resolution — deciding, across messy records from many systems, which ones describe the same person, company or address, and how they relate. Integrating it means mapping source fields into Senzing's feature model and then driving its SDK correctly, and both steps are where projects stall. This official MCP server is aimed squarely at that: it is a build-time assistant for Senzing, not a query interface to your resolved data.

That distinction matters for privacy and is deliberate. The server runs against pre-fetched Senzing documentation and reference material and never reaches into a customer's records, so connecting it exposes nothing. It ships roughly thirteen tools plus a matching set of prompts across six areas: metadata about its own capabilities, data mapping, SDK and code generation, documentation search, troubleshooting and sample data.

The mapping workflow is the centrepiece — a guided eight-step process that profiles your source data, plans the entity structure, maps fields onto Senzing features with fuzzy matching, generates and validates the output, then optionally detects the SDK environment, loads test data, produces a validation report and evaluates the result. SDK tools scaffold working code for Python, Java, C# and Rust covering record ingestion, entity search and redo processing, and can pull real examples from Senzing's GitHub. Troubleshooting takes any of 456-plus error codes and returns causes and resolution steps, with full-text documentation search behind it for deeper context. Three sample datasets — Las Vegas, London and Moscow — are included so an evaluation can run end to end without supplying real data.

It is hosted at https://mcp.senzing.com/mcp over streamable HTTP with no authentication, which makes it one of the easiest MCP servers to try: point Claude, Cursor, VS Code or another client at the URL and ask it to map a CSV.