# opendatalab/MinerU-Ecosystem [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/opendatalab/MinerU-Ecosystem/tree/main/mcp  
**GitHub Stars:** 230  
**Views:** 10  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/opendatalab-mineru-ecosystem

## Description
Official MinerU document parsing MCP (mineru-open-mcp on PyPI). Converts PDFs, doc/docx/ppt/pptx, images, and spreadsheets to Markdown via the MinerU API; free Flash mode without an API key (about 20 pages per file); optional MINERUAPITOKEN for higher limits.

## Tools
Capabilities this server exposes over MCP:

- **parse_documents** — Convert PDF, Office (DOCX, PPTX), spreadsheets (XLSX in Flash mode), images, and http(s) URLs to Markdown using the MinerU cloud API (content is uploaded to mineru.net; do not use for data that must stay on-device). Does not modify source files; may write Markdown under output_dir when saving results.
Auth: without MINERU_API_TOKEN, Flash mode applies (Markdown-only, about 20 pages and 10 MB per file; service rate limits). With MINERU_API_TOKEN, higher limits and optional formats per plan.
Use for extraction and conversion. Use get_ocr_languages only to list OCR language codes, not to parse files. Not for fully offline parsing.
Parameters: file_sources is paths/URLs or objects with source and pages for PDF ranges; language is an OCR code (default ch); enable_ocr defaults to auto (null); set model to html only if every source is a web page URL.
- **get_ocr_languages** — Return supported MinerU OCR and script language codes (e.g. ch, en, japan, latin). Read-only; no uploads. Use before setting the language argument on parse_documents for scanned or multilingual documents. Do not use for file conversion; call parse_documents instead.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "mineru-ecosystem": {
    "command": "uvx",
    "args": ["mineru-open-mcp"],
    "env": {
      "MINERU_API_TOKEN": "",
      "OUTPUT_DIR": ""
    }
  }
}
```

**Requires environment variables:** `MINERU_API_TOKEN`, `OUTPUT_DIR` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What opendatalab/MinerU-Ecosystem MCP server does

opendatalab/MinerU-Ecosystem connects MCP-compatible clients to the official MinerU API at mineru.net. It accepts local file paths and remote URLs, sends the selected documents for parsing, and makes the resulting content available to the client as Markdown or saved output files.

Supported inputs include PDF, DOC, DOCX, PPT, PPTX, and common image formats such as PNG, JPG, JPEG, JP2, WEBP, GIF, and BMP. Flash mode also accepts XLS and XLSX files. Parsed content can retain tables as HTML and formulas as LaTeX. Page ranges can be assigned to individual files, which is useful when processing only selected sections of a document.

## How it works

The server can run over stdio, where the MCP client starts it as a subprocess, or through streamable HTTP for clients that connect to a running endpoint. A typical stdio setup launches the `mineru-open-mcp` PyPI package with `uvx`. For HTTP use, start the process with the streamable HTTP transport and connect to its `/mcp` endpoint.

Single-document results are returned inline when practical. Batch jobs, or responses that are too large for inline delivery, are written to the configured output directory and returned with file metadata, an extraction path, or a download URL. Prompts can also specify a page range, parsing model, OCR language, or output directory.

Files supplied through a client may reside in a sandboxed temporary directory. When parsing local files, provide their full paths so the server can locate them.

## Setup and configuration

Install `uv` and launch the published `mineru-open-mcp` package with `uvx`, or install and run the package directly. The server works without credentials in Flash mode. Set `MINERU_API_TOKEN` to use the full MinerU service and its higher limits. Flash mode is restricted to Markdown output and has lower limits, including an approximately 20-page-per-file limit described for the free path.

`OUTPUT_DIR` controls where batch results and oversized inline responses are saved; its default is `~/mineru-downloads`. An HTTP deployment can be started with a command such as `mineru-open-mcp --transport streamable-http --port 8001`, then registered at `http://127.0.0.1:8001/mcp`.

Documents and URLs submitted for parsing are sent to MinerU. Parsed results are temporarily cached by the service and are not used for training. Local output files are stored in the selected output directory.

## Tools and capabilities

- `parse_documents` parses local files and remote URLs, with per-file page ranges and configurable output handling.
- `get_ocr_languages` lists OCR languages supported by MinerU.
- Flash mode supports PDF, images, DOCX, PPTX, XLS, and XLSX inputs without an API token.
- Authenticated operation supports the broader documented file set and additional output formats.
- Parsing requests can select the pipeline or VLM model, choose an OCR language, and set a custom output directory.

The server is suited to document search, extraction, summarization, and downstream analysis performed by an MCP client after conversion to Markdown.

_Full upstream README: https://allmcps.com/mcp/opendatalab-mineru-ecosystem/readme_