The server itself is a thin single-purpose bridge to Wolfram Alpha. What makes the repository worth cloning is that it ships the rest of the loop — a Gemini client, a Gradio UI, and Dockerfiles for both — so you can watch an LLM consume Wolfram results before wiring anything into your own stack.
An MCP server that fronts the Wolfram Alpha API, so a chat application can send computational queries and get structured knowledge back instead of relying on the model's own recall. The repository also ships the other half of the loop: an MCP client example built on Gemini via LangChain, and a Gradio web interface for switching between Wolfram Alpha, Google AI and query history.
- Math, science and data questions answered by Wolfram Alpha's engine rather than from the model's memory
- A Gradio web interface that switches between Wolfram Alpha, Google AI (Gemini) and past queries
- A command-line client that drives the same server, runnable directly or from a Docker image
- A VS Code entry point: a `.vscode/mcp.json` in your project root, modelled on the template at `configs/vscode_mcp.json`
A Wolfram Alpha AppID in `WOLFRAM_API_KEY`. `GeminiAPI` is only needed for the bundled Gemini client and the Gradio UI, not for the server itself. Install is a clone plus `pip install -r requirements.txt`, or `uv sync`; the pyproject names the package `mcp-wolframalpha` at 0.1.0. Claude Desktop launches it with `python3` against `src/core/server.py`.
One command plus a key — docker run wolframalphaui, then supply credentials