If you're building AI agents that need to understand not just what data exists but what it means, who owns it, and whether it's trustworthy, this is worth looking at. It connects to your OpenMetadata instance via streamable HTTP and exposes the full metadata knowledge graph: technical schemas, data quality signals, lineage, column-level lineage, glossary terms, classifications, metrics, and governance policies across 120+ data sources. Instead of pointing an LLM at raw database tables, you give it governed context so it can reason about business semantics, understand downstream impact, check data freshness, and respect ownership boundaries. Particularly useful if you're already running OpenMetadata and want agents that can navigate your data catalog, answer questions about trust and lineage, or help users discover the right datasets without stepping on governance landmines.
claude mcp add --transport stdio io.github.open-metadata-openmetadata-mcp uvx openmetadata-mcp