About Pdfvision
Pdfvision is an MCP server in the Content category: cLI, MCP server, and TypeScript library that extracts text, layout, and page images from PDFs for AI agents — when extraction goes wrong, it says so per page and says what to do next.
Install
git clone https://github.com/yamadashy/pdfvisionThis server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.
Tools (3)
read_pdfText as Markdown. Without `pages` on a long document it returns a **document map** — page count, outline, per-page quality, warning codes by page range — so an unscoped first call can never blow up the context. `ocr: "jpn+eng"` switches the selected pages to OCR.
Parameters not published.
render_pdfPage or region PNGs as image blocks. Takes a `ref` from an earlier response so you never transcribe coordinates; full-page renders come back with the page's detected visual regions and their refs. Rendering one ref leaves the rest of the set alive, so a search's hits can be rendered one after another.
Parameters not published.
search_pdfOne row per distinct place a hit lands — page, origin, optional context, region, and a short `ref` — with same-source occurrences whose crops resolve to one region — typically repeats within a line or table row — collapsed into a single `×N` row. Names the pages whose native text is unusable, so zero hits is never mistaken for absence.
Parameters not published.
Ask AI
Ask AI about Pdfvision
Powered by Claude · Grounded in docs
Security checks
6 checks have not been run yet.