Vibe Coding Discover

MCP

pdf-mcp

View on GitHub

An MCP server that gives your AI agent agentic RAG over your PDFs, one file or a whole folder: hybrid semantic + keyword search, selective page reads, tables, images, OCR, chart data, and multi-column/CJK layouts. The agent decides when to search; pdf-mcp does the retrieval.

★ 14514 forksPythonMITjztan

MCP server that gives AI agents agentic RAG over one PDF or a whole folder: hybrid BM25 + semantic search, selective page reads, tables, images, chart data, OCR and CJK/multi-column reading order, with a persistent SQLite cache. Ships 13 tools plus STDIO and HTTP transports.

Use Cases

Agentic RAG over a single PDF or a whole folderHybrid keyword + semantic search across document corporaReading only relevant pages to cut context tokensExtracting tables, images and embedded figures from PDFsOCR on scanned PDFs with parallel page processingDigitizing chart data into (x, y) tablesCorrect reading order for multi-column and Japanese tategaki layoutsAdding PDF retrieval to Claude Code, Cursor, VS Code or Codex CLI via MCPServing a shared warm PDF corpus over HTTP transportTriage and per-document overview of large PDF collections

Built With

Language
Python
Frameworks
FastMCP · Model Context Protocol SDK · PyMuPDF · PDFium · fastembed · ONNX Runtime · numpy · Tesseract OCR · Pydantic · SQLite · Docker · Caddy · CUDA

Tags

mcp · mcp-server · agentic-rag · pdf · rag · hybrid-search · semantic-search · bm25 · ocr · table-extraction · chart-extraction · pdf-extraction · cjk · document-processing · sqlite-cache · claude-code