Hi! I am new here, but a long-time data person (25 years doing various kinds of data-driven development). I have created an MCP server that wraps an ML model of my own design with the MongoDB wire protocol to create a sort of “mock” database for development and testing purposes:
The code is open-sourced under the MIT license. You can view a demo video showing how it works with Claude Code here:
This can be useful during design, development, and testing phases when you need “more data” but don’t have more data available. For instance, say you have 100 rows of test data but you really need several million. MongTap, as a generative database, handles that by creating a DataFlood model based on your test data that can then generate thousands of rows pers second for testing purposes.