This is the minimal version: one query tool and a framing prompt, no client, no UI. If all you want is for the model to hand a computational question to Wolfram Alpha instead of guessing, this is the whole surface — set the AppID and go.
A small MCP server that connects to the Wolfram Alpha API. It exposes a single query tool plus a prompt that frames a question for Wolfram Alpha, the way a `!wa` search bang would.
- A question answered by Wolfram Alpha's engine — `query_wolfram_alpha`
- A prompt that wraps a query as "use Wolfram Alpha to answer the following question" — `wa`
A Wolfram Alpha AppID in `WOLFRAM_API_KEY`, which the README marks as required. Launched with `uv --directory <path> run MCP-wolfram-alpha`. The pyproject names the package `mcp-wolfram-alpha` at 0.2.2. For debugging, the author recommends wong2's mcp-cli inspector over the official one because of its environment handling.
One command plus a key — uvx mcp-wolfram-alpha, then supply credentials