Most document tools return whole files or a similarity score. Outlining first, grepping for the pattern, then reading the specific offset is the workflow a person actually uses — and it keeps a long PDF from arriving in a model's context whole. Covering audio and images alongside text in the same type filter is the other useful part: the library is whatever you put in it.
Linkly's endpoint over a personal or team document library. The tool set reads like a filesystem for documents: find, outline, grep, read a slice.
- search — across the library, filtered by document type: pdf, docx, pptx, epub, md, txt, html, image, audio
- outline — the structure of one or more documents, expandable
- grep — a pattern inside a document with configurable context lines before and after
- read — a document by id with an offset, so a long file arrives in slices
- list_libraries, list, explore, find_paths — navigate what exists
- note_save — write a note back
OAuth with a bearer token. Two privacy defaults are worth knowing: LAN access is off unless explicitly enabled with a token, and tunnelling is off unless configured on Pro.
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen