ComPDF MCP
MCP PDF from ComPDF: the agent converts PDFs and images to Word, Excel, Markdown and JSON and merges, splits, protects and compares files via a cloud API
- 98 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- Whole files go to an external cloud API, and contracts and invoices can hold personal and financial data
- The tools encrypt, remove passwords and watermarks, merge and split files, and the result is written to disk
Install
- Works in
Manual install
pip install -e .Run in the repository root inside a virtual environment; then cp .env.example .env and start compdf-streaming-mcp.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Whole files go to an external cloud API, and contracts and invoices can hold personal and financial data
- The tools encrypt, remove passwords and watermarks, merge and split files, and the result is written to disk
This is third-party code. Review the repository files before installing.
What it does
ComPDF MCP is an MCP service for PDF from the ComPDF team. You run it yourself as a Python HTTP service (FastMCP and Starlette), and it sends files to the ComPDF cloud API. The authors' idea is to turn a heavy, noisy PDF or scan into a light structured result first and only then give it to the model, which spends fewer tokens. Tools fall into three groups. Conversion: PDF to Word, Excel, PowerPoint, HTML, RTF, image, CSV, TXT, JSON, Markdown, searchable PDF and OFD, the reverse directions from Word, Excel, TXT, CSV, HTML and images, plus images to Word, Excel, JSON and other formats. PDF operations: merge, split, rotate, delete, insert and extract pages, watermarks, compression, encryption, password removal for authorized files and comparison of two files. Intelligent parsing: document_parse and document_extract. All 44 synchronous tools are available at one /mcp address, and three modules can be connected separately. Long jobs have an asynchronous mode with status checks.
Who it is for. For people who process batches of reports, contracts, invoices and scans and want an agent to convert and edit PDFs before analysis.
Good fit when
- A report or manual must become Markdown or Word before you ask a model for a summary
- Tables from invoices and scanned sheets must be extracted to CSV or JSON
- You need to merge several PDFs, add a watermark and return the final file
- Two versions of a contract must be compared to see the differences
- A scan must become a searchable PDF for retrieval and archiving
Not a fit when
- Documents must not go to a third party: files are sent to the ComPDF cloud API and the service has no local processing
- You need semantic search across a folder of PDFs with context savings: servers like pdf-mcp suit reading and search better
- You want one-command installation from a package manager: the service is deployed from the repository with an .env file
- Files larger than 100 MB decoded: that is the limit for a single file
Example request
Convert report.pdf to Markdown, export all tables separately to CSV, then give a short summary of the report.MCP
- Transport
- http
- Authentication
- API key
Environment variables
COMPDF_API_KEYrequired, secret- Personal ComPDF API key; the client sends it in the X-ComPDF-API-Key header and the server does not store it
Limitations
A personal ComPDF API key is needed; it is free for trials and scenario checks, and larger volumes require contacting sales. The server does not store the key: the client sends it in the X-ComPDF-API-Key header with every request. Accounts from mainland China use a different API address in the settings.
Files travel as Base64, with a limit of 100 MB per file after decoding. Download links for finished files stay valid for 24 hours. Python 3.10 or newer is required, and the service listens on port 8000 by default. For production the docs suggest an HTTPS proxy and access tokens on top of the ComPDF key.
Whether the service is reachable from Russia was not checked for this card.
How to disable. Stop the compdf-streaming-mcp process or run docker compose down and remove the compdf block from the client configuration. The ComPDF key is revoked in the ComPDF Portal account.
FAQ
How do I install the ComPDF MCP PDF server in Claude Code?
Install the service with pip install -e . in a virtual environment, copy .env.example to .env and start compdf-streaming-mcp. Then add a compdf server to ~/.claude.json with type http, the address http://127.0.0.1:8000/mcp and the X-ComPDF-API-Key header taken from the COMPDF_API_KEY variable.
Does ComPDF MCP need a key and what does it cost?
A personal key is needed and is issued for free in the ComPDF Portal after registration. Free access is enough for trials and scenario checks. For large volumes and enterprise deployment the authors suggest contacting sales.
Where do my PDFs go when I use ComPDF MCP?
The service sends files to the ComPDF cloud API. It has no local processing, and links to results stay valid for 24 hours. For confidential documents use this route only if your security policy allows it.
How does ComPDF MCP differ from pdf-mcp and MarkItDown?
ComPDF MCP converts PDFs and images to dozens of formats, edits pages and protects files through a cloud API. pdf-mcp searches and reads PDFs locally, and MarkItDown turns files into Markdown.
Can I connect only some of the ComPDF MCP tools?
Yes. Besides the shared /mcp address with 44 tools there are modules: /mcp/conversion/mcp for conversion, /mcp/ai/mcp for intelligent parsing and /mcp/pdf/mcp for editing and comparison.
Discussion
No questions or reviews yet.
Related
MCP serversAn MCP server that turns PDF, DOCX, XLSX, PPTX, images, audio, web pages and YouTube transcripts into Markdown
MCP PDF for large documents: the agent searches by meaning and keyword across one file or a folder and reads only the pages, tables and scans it needs
Microsoft's utility and MCP server that turn PDF, Word, Excel, PowerPoint, HTML and other files into Markdown for models
Translates scientific PDFs while keeping formulas, tables and layout intact, as a CLI, Docker image and MCP server for agents
An MCP server built on Docling: converts PDFs into structured documents, builds new ones and feeds them to RAG
A PDF-reading MCP server: structured text, tables, OCR, visual crops and page-level citations
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| ComPDF MCP (this entry) | yes | API key | 98 | |
| pdf-mcp | no | not required | 155 | reads and searches PDFs locally without converting anything |
| MarkItDown | yes | not required | 191,675 | turns files into Markdown with a utility on your own computer |
| Citra | yes | not required | 1,050 | an MCP server for reading PDFs with quotes and page numbers |
README in short
The docs explain the idea of processing a document first and giving it to the model second, capability tables for the three groups, the architecture with module addresses, the Base64 file input format, installation on Windows, Linux and macOS, .env settings, ready configurations for Codex, Claude Code, GitHub Copilot in VS Code and Cursor, asynchronous mode with presigned uploads, Docker Compose deployment and example prompts.