Mineru Open MCP Server

@opendatalab

リポジトリ

基本情報

カテゴリ

ファイルとストレージ

トランスポート

stdio

作者

opendatalab

投稿者

莎莎王

Mineru Open MCP Server について

Official MinerU MCP server for document parsing. Convert PDF, DOCX, PPTX, images, and HTML to high-quality Markdown. Supports 109-language OCR, table/formula extraction, and both token-free Flash mode (20 pages / 10MB) and Precision mode (600 pages / 200MB). Works with Claude Des

#high-quality-document-parsing#sota-pdf-extraction#academic-paper-parsing#formula-recognition#table-extraction#full-document-parsing-200-pages-mineru#document-parsing #pdf #ocr #markdown #mcp #claude #cursor #rag #langchain #llamaindex

Mineru Open MCP Server の設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "mineru": {
      "command": "uvx",
      "args": [
        "mineru-open-mcp"
      ],
      "env": {
        "MINERU_API_TOKEN": "your token"
      }
    }
  }
}

Mineru Open MCP Server のツール

Convert PDFs, Word (DOCX), PowerPoint (PPTX), Excel (XLSX in Flash mode), images, and public document URLs—or HTML page URLs—into Markdown using the MinerU cloud API. **World effects:** Reads local paths you pass; fetches http(s) URLs. Uploads file or URL content to mineru.net for processing (do not use for data you must not send off-device). Does not modify or delete originals. May write Markdown (and sometimes images) under `output_dir` or the server default when results are saved or inline content is too large. **Auth & limits:** Without `MINERU_API_TOKEN`, uses free Flash mode: Markdown-only output, service limits apply. With `MINERU_API_TOKEN`, higher per-file page limits and optional extra output formats per MinerU plans; token is read from env (or HTTP Bearer when using streamable HTTP). **Use this when:** The user needs full-document extraction, tables/formulas as HTML/Latexa, batch conversion, or per-file PDF page ranges. **Do not use** for listing supported OCR script codes—call `get_ocr_languages` instead. **Not a substitute** for offline-only or strictly local parsers. **Parameters (intent):** `file_sources` is a list of path/URL strings or `{"source": "…", "pages": "1-5"}` objects (PDF page ranges; Flash allows simple `N` or `N-M`). `enable_ocr` defaults to true. `language` is an OCR/script code (default `ch`); see `get_ocr_languages` for valid values. Set `model` to `"html"` only when every source is a web page URL; otherwise omit. `output_dir` overrides where large or batch results are written.

Return the supported MinerU OCR / script language codes (e.g. ch, en, japan, latin). Read-only; no uploads. Use before setting `language` on `parse_documents` for scanned or multilingual documents. Do not use for converting files—call `parse_documents` instead.

概要

What is Mineru Open Mcp?

Mineru Open Mcp is the official MCP server for the MinerU document parsing API. It converts files such as PDF, DOCX, PPTX, images, and HTML into Markdown, with OCR support for 109 languages and extraction of tables and formulas.

How to use Mineru Open Mcp?

Install by running uvx mineru-open-mcp and configure the MINERU_API_TOKEN environment variable. Add the server definition to your MCP client settings (e.g., Claude Desktop, Cursor, Windsurf). Then use the parse_documents tool to parse files or URLs, get_ocr_languages to list OCR languages, and clean_logs to clean logs (if ENABLE_LOG=true).

Key features of Mineru Open Mcp

Parse PDF, DOCX, PPTX, images, and HTML to Markdown
OCR support for 109 languages
Table and formula extraction in Precision mode
Flash mode (free, no token) for up to 20 pages / 10MB per file
Precision mode (token required) for up to 600 pages / 200MB per file
Works in Claude Desktop, Cursor, Windsurf, and other MCP clients

Use cases of Mineru Open Mcp

Converting scanned PDF documents into editable Markdown with OCR
Extracting tables and formulas from research papers for further analysis
Processing large documents (up to 600 pages) in Precision mode
Batch parsing of multiple file types (DOCX, PPTX, images, HTML) into a unified Markdown format
Integrating document parsing into MCP-compatible AI assistants

よくある質問

What is the difference between Flash mode and Precision mode?

Flash mode is free, requires no token, outputs Markdown only, and is limited to 20 pages or 10MB per file. Precision mode requires an API token, supports up to 600 pages or 200MB per file, and offers richer parsing including table and formula extraction.

What file types are supported for parsing?

The server supports PDF, DOCX, PPTX, images, and HTML files. Input can be local files or URLs.

Do I need an API token to use Mineru Open Mcp?

An API token (`MINERU_API_TOKEN`) is required for Precision mode. Flash mode does not require a token.

How many OCR languages are supported?

OCR supports 109 languages. You can list them using the `get_ocr_languages` tool.

Which MCP clients are compatible?

Mineru Open Mcp works with Claude Desktop, Cursor, Windsurf, and other MCP-compatible clients.

「ファイルとストレージ」の他のコンテンツ

PDF Extraction MCP Server (Claude Code Fork)

xraywu

MCP server to extract contents from a PDF file

Bulk Filesystem Operations MCP Server

strawgate

MCP Server for reading many files at the same time

PDF Reader MCP Server (@sylphlab/pdf-reader-mcp)

sylphlab

📄 The PDF intelligence layer for AI agents — Agent Document Twin, evidence-first extraction, visual crops, OCR provenance, trust reports, and benchmark-gated releases. MCP server for Claude, Cursor, VS Code, and any MCP client.

810