The design choice worth noticing is where the index lives: the docs are cloned and embedded on your machine, so search never goes out to a service and there is no key to manage. The cost is a one-off wait on the first call while that index is built.
A documentation server for IMAGIN.studio — CDN configuration, API references and integration guides. On first run it clones the docs and builds a local vector index; after that it searches that index and refreshes it when the upstream documentation changes.
- The IMAGIN.studio documentation searched by meaning rather than by exact string — `search_docs`
- An index that lives on your machine, so a lookup does not depend on a docs site being reachable
- An index that updates itself when the upstream docs change
No API key and no account — everything runs locally. `uvx imagin-studio-api-docs-mcp`, or `npx -y @imagin.studio/api-docs-mcp` if your client cannot find `uvx`, which is common with GUI apps such as Claude Desktop and Cursor. The first run clones the docs and builds the index, roughly 30 sec. Published as `imagin-studio-api-docs-mcp` on PyPI and `@imagin.studio/api-docs-mcp` on npm, both 0.1.23.
One command — uvx imagin-studio-api-docs-mcp