Install MinerU MCP Server

Pick your client, copy the command, done.

claude mcp add -e "MINERU_API_KEY=${MINERU_API_KEY}" mineru-mcp -- mineru-mcp-server

Required:MINERU_API_KEY+ 4 optional

Environment Variables

Set these before running MinerU MCP Server.

VariableDescriptionRequired

MINERU_API_KEYMinerU API key obtained from the official websiteYes

MINERU_API_BASEBase URL for the remote MinerU APINo

OUTPUT_DIRDirectory path where converted results are savedNo

USE_LOCAL_APIToggle to use a local MinerU API instanceNo

LOCAL_MINERU_API_BASEURL for the local API instance when USE_LOCAL_API is trueNo

Once configured, MinerU MCP Server gives your AI agent access to:

parse_documentsConvert files to Markdown, supporting local paths and URLs with batch processing capabilities.

file_sourcesenable_ocrlanguagepage_ranges

get_ocr_languagesRetrieve the list of languages supported by the OCR engine.

After setup, try these prompts with your AI agent:

→Convert the PDF file located at /documents/report.pdf into Markdown format.

→Convert this URL https://example.com/presentation.pptx to Markdown and enable OCR for English text.

→Get the list of all languages currently supported by the OCR engine.

→Convert the first 5 pages of /data/manual.pdf to Markdown using the Chinese language setting.

Prerequisites & system requirements

An MCP-compatible client (Claude Code, Cursor, Windsurf, Claude Desktop, or Codex)
Python 3.8+ with pip installed
MINERU_API_KEY — MinerU API key obtained from the official website

Alternative installation methods

git clone https://github.com/Tongzhao9417/mineru_mcp.git && cd mineru_mcp && pip install -e .

Conare · memory for coding agents

Save the docs, env vars, and workflow around MinerU MCP Server in Conare so Claude Code, Codex, and Cursor remember it next time.