MinerU Open MCP
Official MinerU MCP server: PDFs, Word, PowerPoint and images become Markdown with formulas and tables
- 229 ★ on the repository
- Last commit
Extended access
Why this level
- The content of your documents is sent to the external mineru.net API and cached temporarily on its servers
- The server writes results to disk in `OUTPUT_DIR` and the `MINERU_API_TOKEN` token ties usage to your account
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/mineru-open-mcpDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The content of your documents is sent to the external mineru.net API and cached temporarily on its servers
- The server writes results to disk in `OUTPUT_DIR` and the `MINERU_API_TOKEN` token ties usage to your account
This is third-party code. Review the repository files before installing.
What it does
MinerU Open MCP opens document parsing through the MinerU Open API to an agent. The `parse_documents` tool takes local files and links to remote ones and returns clean Markdown: formulas go to LaTeX, tables to HTML, columns are assembled in reading order and headers and footers are removed. PDF, images (png, jpg, jp2, webp, gif, bmp), Doc, Docx, Ppt and Pptx are supported, plus xlsx in flash mode. The second tool, `get_ocr_languages`, lists recognition languages. A key is optional: flash mode works at once, free and with no sign-up but returns Markdown only and with lower limits. A token in `MINERU_API_TOKEN` unlocks higher limits and extra formats. For one file Markdown comes right into the chat, and for a batch results are saved to disk in `OUTPUT_DIR` (`~/mineru-downloads` by default) with metadata. Page ranges can be set per file, as well as the model and OCR language. It works over stdio and Streamable HTTP.
Who it is for. People preparing documents for RAG and analysis: analysts, researchers and developers who need exact Markdown from papers and scans.
Good fit when
- An arXiv paper has to become Markdown and its tables explained from a link without downloading
- Only pages 3 to 5 of a PDF are taken to analyze a chapter
- A batch of contracts and reports is parsed with separate page ranges for each file
- A scan in Korean or another language is recognized with the right OCR language
Not a fit when
- Documents are confidential and must not go to an external server: content is sent to the mineru.net API
- You need to work offline: the server calls a cloud API
- You need original layouts and exact typesetting: the output is Markdown and JSON, not a copy of the design
Example request
Parse pages 3 to 5 of this PDF into Markdown, keep formulas in LaTeX and turn the table from the second chapter into a short summaryMCP
- Transport
- stdio, http
- Authentication
- key is optional
Environment variables
MINERU_API_TOKENsecret- MinerU token for higher limits; without it flash mode is used
OUTPUT_DIR- Directory for saved results, ~/mineru-downloads by default
Limitations
Document content is sent to the official MinerU API (mineru.net), results are cached temporarily on their servers and are not used for training by the developers' statement. Flash mode without a token has lower limits and Markdown only. Clients with a sandbox place dragged files into a temporary directory so local files need full paths. A log is written only with `ENABLE_LOG=true`. Running it needs `uv`, the package version in the MCP registry is 1.0.20 and the license is Apache-2.0.
How to disable. Remove the mineru block from the client config (`claude mcp remove mineru`). Clean the results directory `~/mineru-downloads` by hand.
FAQ
Is an API key required?
No, the server starts without a key. Get a token at mineru.net and set it in `MINERU_API_TOKEN` if you need the full set.
How do I connect the server?
Add a block with the `uvx` command and the argument `mineru-open-mcp` to the client, with `env` for `MINERU_API_TOKEN` if needed. For network clients `mineru-open-mcp --transport streamable-http --port 8001`.
Where are results saved?
For batch processing and large output to the `OUTPUT_DIR` directory, `~/mineru-downloads` by default. The reply includes file metadata and a link to the archive.
Which input formats are supported?
PDF, images png, jpg, jpeg, jp2, webp, gif and bmp, plus Doc, Docx, Ppt and Pptx. xlsx files are accepted in flash mode.
What does the MinerU-Ecosystem repository contain?
A command-line tool, SDKs in Python, Go and TypeScript, loaders for LangChain and LlamaIndex, an MCP server and skills for agents such as Claude Code.
Discussion
No questions or reviews yet.
Related
MCP serversA zero-key, zero-dependency skill and CLI that turns PDF, Office and image files into Markdown via MinerU's cloud.
MinerU is a CLI with a document library: search and revisit an already parsed report using its local cache.
An MCP server that turns PDF, DOCX, XLSX, PPTX, images, audio, web pages and YouTube transcripts into Markdown
MinerU PDF Translate turns scientific PDFs into translated PDFs in another language with formulas and figures preserved.
Microsoft's utility and MCP server that turn PDF, Word, Excel, PowerPoint, HTML and other files into Markdown for models
A Rust port of docling that turns PDF, Office, HTML and scans into Markdown and JSON in one binary, with an MCP mode
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| MinerU Open MCP (this card) | yes | key is optional | 229 | |
| MarkItDown | yes | not required | 195,008 | a Microsoft utility for many formats to Markdown rather than MinerU's cloud parsing of formulas, tables and scans |
| MinerU Skill | no | not required | 133 | a keyless skill and CLI for the same PDFs rather than an MCP server with page ranges and model choice |
README in short
The text presents the repository as a toolkit around the MinerU Open API: a command-line tool, SDKs, integrations with LangChain, LlamaIndex, RAGFlow, Dify and FastGPT, an MCP server and skills. Two API modes, precise and quick, are described and the MCP section points to `mcp/README.md`.