mdkit vs Jina Reader
Jina Reader turns URLs into markdown for LLMs; mdkit converts the files you upload — PDF, Office, HTML — into Markdown and JSON. When each fits.
| Jina Reader | A reader API: prefix a URL with r.jina.ai and get the page back as LLM-friendly markdown |
|---|---|
| mdkit | A document conversion API: upload PDF/Office/HTML files, get Markdown or structured JSON back |
Jina Reader is a hosted reader API from Jina AI: you prefix a URL
with https://r.jina.ai/ and it fetches the page and returns LLM-friendly
markdown; it also offers a web-search variant. It is token-metered as part of
the wider Jina AI API platform, with a free allowance to start.
mdkit does not fetch URLs. It converts the documents you already have — you
upload a PDF, Word, PowerPoint, Excel or HTML file to POST /v1/convert (or
through the MCP server) and get clean Markdown back; async jobs add structured
JSON output.
When Jina Reader fits
You want a URL-in, markdown-out primitive — live web content, no file handling, and you're happy inside the Jina API ecosystem. For grounding an LLM on public web pages it is purpose-built.
When mdkit fits
Your inputs are files, not URLs: PDFs from customers, Office documents from a drive, HTML you already fetched with your own client and credentials. One small hosted REST + MCP surface, async jobs with webhooks for batches, a free tier and flat subscription pricing — run by a solo operator, no platform lock-in. Try it free.