The whole server is a live lookup against a public API, which is exactly the case where a model should not be answering from memory. The limit is stated up front rather than discovered: nothing before 2021-05-19 and nothing more than 23 days out, and asking anyway returns an error rather than a plausible guess.
A small MCP server that retrieves Wordle solutions from the Wordle API for a date you name. There is nothing to configure and no credential involved; a Go version of the same server exists separately for a lighter, faster build.
The Wordle solution for a requested date, fetched live from the Wordle API rather than recalled from training data. The available range is fixed by the API: solutions exist from 2021-05-19 up to 23 days into the future, and any date outside that window comes back as an error from the API.
No account and no key. Two install paths, both one config block: Docker, which the author recommends — `ghcr.io/cr2007/mcp-wordle-python:latest`, run with `docker run --rm -i --init`, and worth pulling once first if your client errors on the first launch — or `uvx`, which needs [`uv`](https://docs.astral.sh/uv) installed and runs the package straight from the git URL as `mcp-wordle`. Releases are cut from version tags, and the publish workflow rejects a release whose tag disagrees with the project version or whose uvx and Docker checks fail.
One command — docker pull ghcr.io/cr2007/mcp-wordle-python:latest