Skip to content
MCP server

MinerU MCP

Parse PDFs, Office files and images to Markdown with tables and LaTeX, self-hosted

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
self-hosted, CLI, Linux
Runs
Locally
Version
0.1.0
Listing updated
today

Similar MCP servers

About MinerU MCP

MinerU MCP is an open-source MCP server for files and documents. It runs locally on your machine and is started by your AI client.

This server is a thin MCP adapter in front of a MinerU API that you run yourself. The agent sends a document by local path or as Base64 and gets back Markdown with headings, HTML tables and LaTeX equations; the heavy parsing models stay on your own hardware and the MCP client installs nothing like PyTorch. The README is written in Spanish with an English version.

parse_document runs synchronously, typically 30 to 120 seconds, for PDF, DOCX, PPTX, XLSX or images. submit_parse_task starts an asynchronous job for documents over about 50 pages, with get_task_status and get_task_result to follow it, and mineru_health reports whether the backend is up, asleep from inactivity or broken. Options include backend (pipeline needs about 1.5 GB of VRAM, hybrid-engine and vlm-engine about 4 GB), parse method, OCR language, formula and table switches and zero-based page ranges. With hybrid-engine, high effort and image analysis, figures can be turned into Mermaid diagrams.

It runs over HTTP on port 8202 for a shared server, with an optional bearer token, or over stdio through Docker or uvx. For privacy it only sends documents to a loopback or host.docker.internal address and refuses any other MinerU URL. It does not include mineru-api or the models, and can start mineru-api through systemd if it is stopped.

Install

Run in a terminal

claude mcp add mineru-mcp -- docker run -i --rm ghcr.io/germaniu/mineru-mcp:0.1.0

Add to claude_desktop_config.json

{
  "mcpServers": {
    "mineru-mcp": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/germaniu/mineru-mcp:0.1.0"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "mineru-mcp": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/germaniu/mineru-mcp:0.1.0"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "mineru-mcp": {
      "type": "stdio",
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/germaniu/mineru-mcp:0.1.0"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "mineru-mcp": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/germaniu/mineru-mcp:0.1.0"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "mineru-mcp": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/germaniu/mineru-mcp:0.1.0"
      ]
    }
  }
}

Using several servers? Build one config file with the MCP config generator.

Configuration

  • MINERU_URLURL of the mineru-api engine. Only loopback or host.docker.internal is accepted; any other host is treated as external and documents are not sent.
  • MINERU_BACKENDBackend used when a call leaves backend empty. pipeline needs the least VRAM; hybrid-engine and vlm-engine need ~4GB.
  • MINERU_PARSE_METHODExtraction method used when a call leaves parse_method empty.
  • MINERU_LANGOCR language used when a call leaves lang empty.

Tools this server provides (5)

Taken from the project documentation; the running server may differ.

parse_document
Synchronous parse of a PDF, DOCX, PPTX, XLSX or image to Markdown.
submit_parse_task
Starts an asynchronous parse and returns a task_id.
get_task_status
Returns pending, processing, completed or failed.
get_task_result
Returns the Markdown of a completed task.
mineru_health
Reports availability, version and queued tasks of mineru-api.

What it does

Document to Markdown
Converts PDF, DOCX, PPTX, XLSX and images with tables and LaTeX.
Async parsing
submit_parse_task and task tools handle long documents.
Selectable backends
Choose pipeline, hybrid-engine or vlm-engine by VRAM and accuracy.
Page ranges
Parse only chosen pages, counted from 0.
Local-only guard
Refuses to send documents to a non-loopback MinerU host.
Health check
Distinguishes a sleeping backend from a broken one.

Good for

  • Convert research papers with equations into RAG-ready Markdown
  • Extract tables from a long report asynchronously
  • Share one GPU-backed parser across a team of agents
  • Parse a few pages of a large PDF
  • Turn figures into Mermaid diagrams

Pricing

Open source

Free and self-hosted

MinerU MCP FAQ

Does it include the MinerU engine?

No. mineru-api and the models are installed separately on the host.

Are documents sent to the cloud?

No. Only loopback or host.docker.internal MinerU URLs are accepted.

How much GPU memory does it need?

About 1.5 GB for the pipeline backend and about 4 GB for hybrid-engine or vlm-engine.

How do I run it?

Over HTTP on port 8202, or over stdio with Docker or uvx.

Is MinerU MCP free?

Yes. MinerU MCP is open source, so you can use and self-host it for free.

How do I add MinerU MCP to Claude Code?

Run this command in your terminal: claude mcp add mineru-mcp -- docker run -i --rm ghcr.io/germaniu/mineru-mcp:0.1.0

What are the best alternatives to MinerU MCP?

The closest alternatives to MinerU MCP are SSRS2, Csv, Google_sheets, Onedrive and Zovo Spreadsheet Compare.

Where this comes from

  • Description: written from the project's README.
  • Install steps: from the official MCP Registry entry.

How AIPoweredBot checks servers