Convert documents with MinerU inside Claude Code

Install

npx skills add tanis90/pdf-converter-mineru

Who it is for

If you already handle local documents in Claude Code and want to turn PDFs, images, or Office files into editable text, this skill fits. Officially it is a Claude Code skill on the MinerU ecosystem for converting PDFs and other documents. The MinerU page is

https://mineru.net/

Output formats are written as Markdown, Word, HTML, LaTeX, or plain text. The named environment is Claude Code: when you ask to convert or extract a document, Claude calls mineru-open-api automatically. The page gives no install commands for other agents and does not write support as an exclusive roster.

The better fit has a PDF or office file that can be converted locally, and wants to see the default flash-extract result first, not jump into full extract. Official notes:

https://github.com/tanis90/pdf-converter-mineru

First try in 3 minutes

Use only this official install path:

npx skills add tanis90/pdf-converter-mineru

Prerequisite: you can already open Claude Code locally and have a local PDF or office document ready. Install is that command; the page gives no second installer.

Pick a low-risk conversion for the first task. First confirm with Claude the source path, the target format (start with Markdown or plain text), and where the result should be written. After you agree, let it convert. The default mode is flash-extract; no auth is required first. Conversion writes a new file and is also an outbound call. Confirm the path before running.

Success signal: after the skill is installed, Claude automatically uses mineru-open-api and produces an openable Markdown or plain-text file, not a “not installed” prompt. If it wants mineru-open-api auth, stop first and confirm that you mean to unlock extract mode. That is not required for the first try.

Use cases

Case 1: Convert a PDF to Markdown with default flash-extract.

Officially a PDF can become Markdown, and the default is flash-extract. The first time, convert one known PDF and check that titles and paragraphs are still there. The page does not say what happens with headers, footers, scans, or encrypted PDFs, nor the quality gap between flash-extract and extract. What you can confirm is only the default mode name, and that Claude calls mineru-open-api automatically.

Case 2: Authenticate first, then extract an image or Office document.

Officially, running mineru-open-api auth unlocks extract mode. Conversion covers images and Office documents; targets can still be Markdown, Word, HTML, LaTeX, or plain text. Run auth alone first, and only after you see unlock success pick an image or office file. Auth changes local credentials; confirm the command window comes from mineru-open-api auth before continuing. The page does not list the auth fields or extra extract switches.

Both cases sit on the same npx install command. Do not try other install methods in parallel the first time.

Limits

The only copyable official install command is npx skills add tanis90/pdf-converter-mineru. The default is always flash-extract. For extract you must first finish mineru-open-api auth successfully. The page does not make auth a prerequisite of the first conversion, and it does not say you can skip Claude and use this as a standalone converter.

Coverage is PDF, images, and Office into five text or document formats, but no file-size, page-count, or language range is given. Conversion writes new files; confirm the path first. MinerU is written as the underlying ecosystem; details follow