Secoda MCP

Secoda's MCP server gives an AI client access to a data catalog and the warehouse behind it, which lets an agent answer a data question with both context and numbers. search_data_assets finds tables, columns, charts and dashboards, while the broader search tool covers every Secoda resource including documentation. retrieve_entity returns the full detail for a specific asset, entity_lineage walks upstream and downstream dependencies so an agent can tell what breaks if a table changes, and glossary returns business term definitions so a metric name is interpreted the way the organization defines it rather than the way a model guesses. Two more tools round out the surface. run_sql executes queries against connected data warehouses, turning a cataloged table into actual results in the same conversation. chart works with charts and visualizations, and get_secoda_docs retrieves Secoda documentation and knowledge articles for questions about the platform itself. The server is a Python process configured through client config files for clients such as Cursor and Claude Desktop, authenticated with an API token generated in workspace settings, with an optional API URL for non default deployments. All operations respect the existing governance and access controls of the workspace, so an agent inherits the permissions of the token rather than bypassing them. It is aimed at data teams that already run Secoda as their catalog and want lineage aware answers, and at analysts who would otherwise switch between a catalog UI and a SQL editor to answer one question.