HubDB rows are usually edited through HubSpot's interface and read through templates, which makes them awkward to interrogate. Exposed as SQL tables they become answerable — how many rows match this, which entries are missing a field — in one question rather than by exporting and filtering. This build is read-only with no write tool present; CData sells a separate MCP server with full read-write if you need to change rows.
A read-only MCP server from CData that puts HubDB behind a SQL interface. The CData JDBC driver models HubDB as relational tables; this server wraps it so an assistant can list the tables, read their columns, and run SELECT queries against live data without you writing SQL.
- The HubDB tables available, returned as CSV with a header row — `get_tables`
- The columns on a table, so the query matches the real schema — `get_columns`
- A SQL SELECT executed against live HubDB content — `run_query`
- A configurable prefix on the tool names, so this can sit alongside other CData servers in the same client config
- A `Tables` setting that limits which tables are reachable at all, rather than exposing everything by default
A licensed CData JDBC Driver for HubDB, installed and licensed separately, plus Java. Build from a clone with `mvn clean install` to produce `CDataMCP-jar-with-dependencies.jar`. Connection settings live in a `.prp` file: prefix, server name and version, the path to `cdata.jdbc.hubdb.jar`, the driver class `cdata.jdbc.hubdb.HubDBDriver`, the JDBC connection string from the driver's utility, and an optional table list. Transport is stdio, so client and server run on the same machine.
Build from source — clone the repository and build it, then point your client at the binary