Collibra MCP

Collibra's MCP server connects AI assistants to its data intelligence platform so an agent works from governed context instead of raw table names. Tools let it discover data assets across communities and domains, read asset attributes, relations and certification status, trace upstream and downstream lineage, search the Business Glossary for approved terms and definitions, and surface data products along with the people accountable for them. Collibra's MCP server connects AI assistants to its data intelligence platform so an agent works from governed context instead of raw table names. Tools let it discover data assets across communities and domains, read asset attributes, relations and certification status, trace upstream and downstream lineage, search the Business Glossary for approved terms and definitions, and surface data products along with the people accountable for them. Tool discovery happens dynamically at runtime, which means the operations an agent sees reflect what your instance exposes and what your account is allowed to do rather than a fixed list baked into a client config. The transport is streamable HTTP and authentication runs over OAuth, so calls execute under the signed-in user's Collibra permissions. An assistant cannot read an asset the person behind it could not open in the interface, which is the property governance teams care about most when opening a catalog to automation. Collibra distributes the server for direct connection and through cloud marketplaces such as Databricks, making it straightforward to place next to the compute where analysis already happens. It is built for governance leads, stewards and analytics teams that treat Collibra as the source of truth for definitions and want those same guardrails applied when an agent answers a question about the data.