doc parse mineru selfhosted

    1

    Self-hosted MinerU recipe & agent skill — read PDF, Office, scanned (OCR), EPUB and HTML docs through your own MinerU server.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    doc parse mineru selfhosted

    Example session with this skill installed

    Parse the file 'q3_report.pdf' pages 1-5 using the advanced tier on our local server and extract the revenue table.

    • Read your context and instructions
    • Compiled the doc parse mineru

    I have submitted the parsing request for pages 1-5 of q3_report.pdf to the local MinerU server using the advanced tier. The document was successfully processed, and I have extracted the primary financial data from page 3.

    Extracted Financial Data (q3_report.pdf)

    MetricQ3 2023 (Actual)Q3 2024 (Projected)Variance (%)
    Total Revenue$4.2M$4.8M+14.2%
    Cost of Goods$1.1M$1.3M+18.1%
    Operating Margin24%26%+8.3%
    Net Income$840K$912K+8.5%

    The advanced tier successfully identified the multi-column layout on page 3 and reconstructed the table structure. I also performed a flash tier check to verify the raw numeric values, which match the layout-preserved version above.

    Next steps

    1. Run mineru search "Year-over-Year" to find related historical context in previously parsed reports.
    2. Use mineru read "doc:ab12cd3/tier:advanced/page:3/block:12" --format image if you need to visually verify the chart accompanying this table.
    3. Export the full five-page summary to Markdown using mineru parse "q3_report.pdf" --tier advanced --output ./q3_summary.md.
    Extracted Financial Data (q3_report.pdf

    Connects securely to your tools. The creator never sees your data.

    About this skill

    NOTICE: The skill instructions and deployment guide are currently written in Chinese.

    Give your agent a scanned report, ask it to extract a table, and have it check the figures against the original page. This free skill pairs a document-reading workflow with a deployment recipe for MinerU (>=4.2.2). Your agent uses your own machine and GPU server for parsing, so your documents stay on your infrastructure.

    What your agent can do

    • Read complex PDFs and scans. The skill runs Flash locally and Advanced on your GPU server in parallel. Flash provides a text extraction for comparison, while Advanced reconstructs complex layouts. When the results disagree on a number, label, or unit, the agent can inspect the original page.
    • Read native documents locally. Office files, HTML, EPUB, OFD, CSV, and TSV go through Flash on the client machine. Embedded images in these formats do not receive OCR.
    • Return to the source. The agent can search previously parsed content, reopen a specific page or block across sessions, and export selected PDF pages as Markdown. Requesting more pages only parses the missing ones.

    What’s included

    The repository includes an Agent Skills–compatible skill and a deployment guide covering model installation, GPU server setup, systemd startup, and client configuration. You’ll need the MinerU CLI on the agent’s machine and a self-hosted GPU server for Advanced parsing.

    Get the free skill and deployment guide on GitHub

    Does not OCR embedded images in Office/HTML formats. PDF range defaults to 10 pages. Requires pre-configured self-hosted MinerU GPU server.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 day ago

    • Free with an Agensi account

    Permissions

    Terminal / Shell
    Read Files
    Listed1 day ago

    What's inside

    Frequently Asked Questions

    This skill acts as an orchestration layer that automates the complex CLI commands and server management required to run MinerU. It handles job submission, status monitoring, and layout verification, turning raw GPU power into a structured API-like experience for your agent.

    You will need a self-hosted GPU server with MinerU v4.2.2 installed and accessible via local CLI. The skill assumes your environment meets MinerU's hardware requirements for VLM and OCR processing.

    Yes, it is designed for maximum privacy by processing all files on your local or private GPU infrastructure. No document data is sent to external cloud APIs, making it ideal for sensitive financial or legal documents.

    The 'Advanced' tier uses Vision Language Models (VLM) for high-fidelity layout and table extraction, while the 'Flash' tier prioritizes speed for simpler documents. This skill intelligently handles both, allowing for cross-tier verification to ensure accuracy.

    The skill includes the logic for multi-format parsing (PDF, Office, etc.), job status orchestration, persistent caching, and smart navigation locators. Please note that the MinerU engine and GPU drivers must be installed separately on your host.

    New in LLM Ops