
Pdf Mcp
Gives Claude surgical access to PDFs through eight specialized tools built on PyMuPDF. The workflow is efficient: call pdf_info to get page count and TOC, use pdf_search for hybrid BM25+semantic search with paragraph excerpts, then pdf_read_pages for targeted reads or pdf_render_pages to feed diagrams to vision models. Includes OCR via Tesseract for scanned documents, structured table and image extraction, and SQLite caching that persists text, embeddings, and rendered pages across server restarts. Reach for this when you need to interrogate large documents without dumping entire PDFs into context. HTTPS-only URL fetching with SSRF protection blocks local network access.
- GitHub stars
- 168
- tools
- 13
- transport
- STDIO
- env vars
- 3
- last updated
- Nov 7, 2026
Tools
13 toolsPublic tool metadata for what this MCP can expose to an agent.
pdf_infoPage count, metadata, TOC summary, scanned-page detection. Call first.
Page count, metadata, TOC summary, scanned-page detection. Call first.
No parameter schema in public metadata yet.
pdf_searchHybrid search (keyword + semantic) by page or section; hits carry an excerpt, source coordinates, the enclosing outline section and a table's lead-in sentence
Hybrid search (keyword + semantic) by page or section; hits carry an excerpt, source coordinates, the enclosing outline section and a table's lead-in sentence
No parameter schema in public metadata yet.
pdf_read_pagesRead specific pages or ranges, with OCR on demand, tables, and embedded images
Read specific pages or ranges, with OCR on demand, tables, and embedded images
No parameter schema in public metadata yet.
pdf_read_allRead a whole document in one call, byte-capped
Read a whole document in one call, byte-capped
No parameter schema in public metadata yet.
pdf_get_tocFull table of contents for documents with many bookmarks
Full table of contents for documents with many bookmarks
No parameter schema in public metadata yet.
pdf_render_pagesRender pages as PNG for vision models: diagrams, handwriting, scans
Render pages as PNG for vision models: diagrams, handwriting, scans
No parameter schema in public metadata yet.
pdf_extract_chartChart data as exact (x, y) tables, read from plot geometry
Chart data as exact (x, y) tables, read from plot geometry
No parameter schema in public metadata yet.
pdf_corpus_warmWarm a folder of PDFs into the cache within a time budget
Warm a folder of PDFs into the cache within a time budget
No parameter schema in public metadata yet.
pdf_corpus_overviewPer-document triage cards for a folder
Per-document triage cards for a folder
No parameter schema in public metadata yet.
pdf_corpus_searchSearch across a folder; hits carry document and page provenance
Search across a folder; hits carry document and page provenance
No parameter schema in public metadata yet.
pdf_cache_statsPer-document cache breakdown and total size
Per-document cache breakdown and total size
No parameter schema in public metadata yet.
pdf_cache_clearClear expired or all cache entries
Clear expired or all cache entries
No parameter schema in public metadata yet.
server_infoWhich optional features and config are active
Which optional features and config are active
No parameter schema in public metadata yet.
Sponsors
Configuration
PDF_MCP_CACHE_DIRDirectory for storing PDF cache (default: ~/.cache/pdf-mcp)
PDF_MCP_CACHE_TTLOptional time expiry in hours; unset keeps a document while its PDF exists
PDF_MCP_CACHE_MAX_MBCache size cap in MB, least recently used evicted first (default: 2048; 0 = no cap)
Readme
pdf-mcp
Agentic RAG over your PDFs, with the evidence to check every answer.
Details
- Repository
- github.com/jztan/pdf-mcp
- Categories
- AI & LLM Tools, Documents & Knowledge, Search & Web Crawling
- Package
pdf-mcp- Transport
- STDIO
- Registry
- MCP registry, active