Pdf Tools MCP

What danielkennedy1/pdf-tools-mcp MCP server does

The danielkennedy1/pdf-tools-mcp MCP server provides PDF utilities for agents using the Model Context Protocol. Its local operations work with PDFs available to the server, while remote operations can render pages from a remote PDF or download a PDF from a URL into local storage.

The toolset covers both inspection and document manipulation. Agents can render pages as images, read document metadata, extract text blocks and detailed text information, create page snippets, merge pages into one long page, and fuse pages from multiple documents into a new document. The available features make it suitable for workflows that need to inspect PDFs visually or structurally before performing a limited set of transformations.

How it works

The server registers PDF operations as MCP tools and exposes them through the MCP protocol. A compatible client sends a tool request, and the local Python process performs the requested PDF operation using the project’s PDF-processing dependencies.

Remote workflows include fetching a PDF from a URL for local use and rendering pages from a remote document. Local documents are stored using UUID4 filenames, according to the README, which avoids using predictable filenames for stored PDF files. The project requires Python 3.12 or newer.

Rendering is available for both local and remote PDFs. Text extraction returns text blocks and more detailed text information, while metadata extraction is intended for document inspection. Page-oriented operations include freeform or full-width snippets, page merging, and combining pages from several documents.

Setup and configuration

The documented manual setup starts by cloning the repository and installing the project in editable mode with uv pip install -e .. After installation, start the server with:

python -m src.main

The README also documents Smithery installation for Claude Desktop using the Smithery CLI. That route is client-specific and is separate from the manual Python setup. No API keys, environment variables, or external service credentials are specified in the provided material.

The project lists PyMuPDF, aiohttp, aiofiles, python-magic-bin, and the Python MCP framework among its dependencies. These dependencies support PDF processing, asynchronous file and network operations, file-type detection, and MCP tool registration.

Tools and capabilities

The danielkennedy1/pdf-tools-mcp MCP server supports these documented capabilities:

• Render pages from local PDFs as images.
• Render pages from remote PDFs as images.
• Extract PDF metadata.
• Extract text blocks and detailed text information.
• Create freeform or full-width page snippets.
• Merge multiple pages into one long page.
• Fuse pages from multiple documents into a single document.
• Download PDFs from URLs to local storage.

Limitations and notes

The provided documentation describes PDF rendering, extraction, downloading, and page/document operations, but does not specify supported PDF versions, output image formats, page-range syntax, maximum file sizes, or error behavior. It also does not list named MCP tool identifiers or detailed request and response schemas.

The README states that Python 3.12 or newer is required. Because the server performs remote downloads, its runtime needs access to the relevant URLs, but the provided material does not define network restrictions or timeout settings. License information identifies the project as MIT-licensed.