Pdf Mcp
Let Claude Code or Cursor extract, search, and reuse PDF content in agent workflows without re-parsing the same files every turn.
Overview
io.github.jztan/pdf-mcp is an MCP server for the Build phase that processes PDFs for AI agents with an intelligent on-disk cache.
What is this MCP server?
- Production-oriented PDF processing exposed as MCP tools over stdio
- Intelligent file cache with configurable directory and TTL (default 24 hours)
- PyPI package pdf-mcp v1.9.0 with PDF_MCP_CACHE_DIR and PDF_MCP_CACHE_TTL env vars
- Fits local-first agent stacks that repeatedly query manuals, specs, or legal PDFs
- Server version 1.9.0 on PyPI identifier pdf-mcp
- Default cache TTL 24 hours via PDF_MCP_CACHE_TTL
- Default cache directory ~/.cache/pdf-mcp when unset
Community signal: 55 GitHub stars.
What problem does it solve?
Agents waste context and time re-opening the same PDFs because there is no cached, tool-backed PDF layer in the MCP client.
Who is it for?
Indie builders who pipe contracts, specs, or research PDFs into Claude Code or Cursor daily and want server-side caching without a custom script.
Skip if: Teams that only need one-off human PDF editing in a desktop app with no agent integration.
What do I get? / Deliverables
After you register pdf-mcp, your agent can call PDF tools with cache-aware reuse so document workflows stay fast and repeatable.
- Registered stdio MCP server for PDF operations
- Configurable on-disk PDF cache for repeated agent calls
- Tool surface agents can invoke instead of raw file pastes
Recommended MCP Servers
Journey fit
PDF ingestion sits in Build because solo builders wire document tooling into agents while shipping features, not during idea research alone. Agent-tooling is the canonical shelf: this is an MCP bridge for LLM runtimes, not a standalone CLI product.
How it compares
MCP PDF integration with caching, not a browser automation or markdown-only docs skill.
Common Questions / FAQ
Who is pdf-mcp for?
Solo and indie builders using MCP-enabled coding agents who need repeatable PDF read and process operations during product work.
When should I use pdf-mcp?
Use it during Build when your agent workflow depends on PDFs and you want cached MCP tools instead of uploading the same file every session.
How do I add pdf-mcp to my agent?
Install the PyPI package pdf-mcp, set PDF_MCP_CACHE_DIR or PDF_MCP_CACHE_TTL if needed, and add a stdio MCP server entry in your client config pointing at the published server.