Two things follow from reading Dash's own databases. Documentation you have downloaded is available with no network and no rate limit, and the answer matches the docset version on your disk rather than whatever the model absorbed. The one thing to know going in is that `search_docs` wants an exact name, not keywords — `list_frameworks` and `list_entries` are how you find the name when you do not have it.
A bridge between Dash and an AI client. It queries the docset databases on disk directly, so documentation lookups happen locally and instantly rather than as a search-and-scrape.
- `search_docs` searches by exact name inside a docset, with an optional language filter, ranked exact before prefix before substring
- `list_available_docsets` and `list_available_cheatsheets` show what is actually installed on your machine
- `list_frameworks` lists frameworks and types inside a docset; `list_types` lists the entry kinds — Class, Protocol, Function and so on
- `list_entries` filters entries by type and name prefix; `list_languages` and `list_docsets_by_language` cross-reference by language
- `search_cheatsheet` searches inside a cheatsheet; `list_cheatsheet_categories` and `fetch_cheatsheet` browse or pull one whole
- Eleven tools in all, over 165+ supported docsets, with framework and class results carrying drilldown notes for exploring members
macOS, since Dash is Mac-only, with Dash installed and the docsets you want already downloaded. Python 3.10 or higher and the uv package manager. Run it as `uvx docsetmcp` with no install step, or `pip install docsetmcp`. DOCSET_PATH and CHEATSHEET_PATH override the default Dash directories, and `--additional-docset-paths` searches several locations at once.
One command — pip install docsetmcp