Pdf MCP
Agentic RAG over your PDFs, one file or a whole folder, as a single MCP tool. Its 13 documented tools cover pdf, corpus, pages, cache. It runs locally over stdio via the published package.
People who work with pdf, corpus and pages and want it reachable from Claude, Cursor, VS Code, or another MCP client. The project is written in Python.
VERIFIED ACTIVE
LAST COMMIT 2026-09-17 · ★ 135 · #2 OF 41 MAINTAINED FILESYSTEM · VERIFIED 2026-09-18
MIT · Python servers · how we verify → /methodology
01 · Install Pdf MCP
before you install - you'll need
Set PDF_MCP_AUTH_TOKEN before connecting.
Claude Code
claude mcp add jztan-pdf-mcp -- uvx pdf-mcp Claude Desktop / Cursor / VS Code - add to config
{
"mcpServers": {
"jztan-pdf-mcp": {
"command": "uvx",
"args": [
"pdf-mcp"
]
}
}
} Same JSON for Cursor. For VS Code, rename the top-level key from `mcpServers` to `servers`.
Using another client? Same JSON, different key
Claude Desktop · mcpServers
Cursor · mcpServers
VS Code · servers
Windsurf · mcpServers
Zed · context_servers
Cline · mcpServers
Roo Code · mcpServers
Continue · mcpServers
LibreChat · mcpServers
Gemini CLI · mcpServers
Codex CLI · mcp_servers
Full setup guides: every client.
02 · Evidence
Security posture
What to check before giving this server access to your agent - from the registry, GitHub, and our own probes. We don't score safety; we show what's verifiable.
runs as local process (stdio) - runs on your machine with your user's permissions
license MIT - declared in the repository
pypi package pdf-mcp - check the name against the project README before installing (PyPI has no namespace ownership)
registry namespace io.github.jztan is GitHub-verified and matches the repo owner
03 · What Pdf MCP can do
Prose above is summarized from the project's README and registry record - no invented capabilities.
What you can build
With this server connected, an agent can read specific pages or ranges, with OCR on demand, tables, and embedded images, read a whole document in one call, byte-capped, and search across a folder, with document and page provenance.
Capability map
Tools grouped from the project's README - what Pdf MCP lets an agent do.
12 tools - e.g. Page count, metadata, TOC summary, scanned-page detection. Call first.; Hybrid search (keyword + semantic), page or section granularity, paragraph or context-window excerpts with source coordinates; Read…
pdf_info · pdf_search · pdf_read_pages · pdf_read_all · pdf_get_toc · pdf_render_pages…
Server
1 tool - e.g. Which optional features and config are active
server_info
Latest releases
v3.2.0 · 2026-09-12
Highlights · Embedding, the step behind semantic search and corpus warming, is faster on CPU in this release and can now run on an NVIDIA GPU. · On CPU, embedding 778 real page units on an M4 Pro went from 25.9 s to…
v3.1.0 · 2026-09-05
Highlights · Search results now tell you when a document has no searchable text. A scanned, never-OCR'd PDF used to return zero hits that looked identical to "term not found". Every search and warm response now carries…
v3.0.0 · 2026-08-29
Highlights · pdf-mcp 3.0.0 replaces the runtime PDF engine with a permissive, fully MIT-compatible stack, and quality was gated on the full benchmark set before the swap. Excerpt containment rose to 0.773 from 0.707…
04 · Who maintains Pdf MCP
pdf-mcp is maintained by jztan. We track 3 MCP servers from jztan - 3 actively maintained, 215 combined GitHub stars, oldest repo from May 2025. Full record: all servers from jztan.
05 · Facts
- repository
- github.com/jztan/pdf-mcp
- category
- filesystem - ranked #2 of 41 actively-maintained filesystem servers as of 2026-09-18.
- release cadence
- 10+ releases in the last 90 days (latest 2026-09-12)
- registry
- io.github.jztan/pdf-mcp (active, first published 2026-02-01 · 28 versions)
- packages
- pypi:pdf-mcp
06 · Pdf MCP FAQ
Is Pdf MCP still maintained?
Yes - as of 2026-09-18, its last commit was 2026-09-17 and it shipped 10+ releases in the last 90 days. We re-verify nightly.
What can Pdf MCP do?
With this server connected, an agent can read specific pages or ranges, with OCR on demand, tables, and embedded images, read a whole document in one call, byte-capped, and search across a folder, with document and page provenance.
How do I install Pdf MCP?
Run `uvx pdf-mcp`. The README documents one environment variable (PDF_MCP_AUTH_TOKEN) to set first. Set PDF_MCP_AUTH_TOKEN before connecting. You can also paste the ready-made client config above.
Does Pdf MCP run locally?
Yes - it's a stdio server: it runs on your machine (via uvx) with your user's permissions. Your data stays local unless the server itself calls external APIs.
07 · Alternatives to Pdf MCP
Alternatives to Pdf MCP
Maintained filesystem servers if Pdf MCP isn't the fit.
- Server Filesystem MCP server for filesystem access ★ 39,038 · 2026-09-11
- Directory A directory of product launch platforms: where your product qualifies and how each form is filled. ★ 297 · 2026-08-19
- Remote Filesystem MCP server for remote filesystem operations on cloud storage (Google Cloud Storage). ★ 80 · 2026-08-30
- Keibi Drop Give two agents on two machines the same encrypted folder. P2P, no cloud, only read bytes move. ★ 60 · 2026-09-16
- V8 Run JavaScript and TypeScript in sandboxed V8 isolates with policy-controlled host capabilities. ★ 57 · 2026-09-08
- Work OS Manage your WorkOS workspace in plain language: orgs, users, SSO, Directory Sync, and AuthKit. ★ 46 · 2026-09-17
More filesystem MCP servers · Virtual Filesystem · Cairn · Devtools · Octofs
More Python MCP servers · Qt4 Doc MCP Server · Redmine MCP Server · Deepmiro · Whatsapp Macos · Unifi Applications · see all