Only SELECT gets through, which makes this safe to point at a live account but means it cannot create or update anything — CData directs you to its own commercial MCP server for that. The setup cost is front-loaded and mostly belongs to the driver rather than the server: installing it, licensing it, and getting a connection string that tests clean. Once that works, the MCP side is a JAR and a properties file.
A thin MCP wrapper around the CData JDBC Driver for Pinterest. The driver models the Pinterest API as relational tables, and the server exposes those tables so an assistant can answer questions with SELECT statements it writes itself.
- Three tools, each named with the prefix you set in the connection properties file: list the tables the driver exposes, list a table's columns, and run a SQL SELECT query
- Results come back as CSV with a header row on the first line
- The Tables setting controls scope — leave it blank for everything, or name the tables you are willing to expose
- The same server pattern covers CData's other JDBC sources, so the shape carries across connectors
Java and Maven — `mvn clean install` builds a JAR with dependencies. Separately, the CData JDBC Driver for Pinterest must be installed and licensed by running its jar with `--license` and entering a name, an email and either TRIAL or a licence key. Then write a .prp file naming the prefix, the driver path, the driver class and the JDBC connection string; the driver ships a utility that builds and tests that string, opening a browser for OAuth. Transport is stdio only, 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