# Read any document URL as Markdown

Put https://reducto.ai/read/ in front of any public document URL and GET it: the response is the document as Markdown (text/markdown), parsed by Reducto with the r-1 mini model, tables kept. Without a key it reads the first 5 pages, a few times per day per IP; with Authorization: Bearer <REDUCTO_API_KEY> it reads the whole document on your own key and limits. It works for PDF, DOCX, PPTX, XLSX, PNG and JPG.

## Examples

curl:

```
curl https://reducto.ai/read/https://arxiv.org/pdf/1706.03762
```

With your key:

```
curl https://reducto.ai/read/https://example.com/annual-report.pdf \
  -H "Authorization: Bearer $REDUCTO_API_KEY"
```

JSON:

```
curl https://reducto.ai/read/https://example.com/deck.pptx \
  -H "Accept: application/json"
# {"source": "...", "markdown": "...", "pages_processed": 5, "mock": false, "next": {...}}
```

Python:

```
import requests

md = requests.get("https://reducto.ai/read/https://example.com/contract.docx", timeout=120).text
print(md[:500])
```

## Limits and rules

- Without a key: first 5 pages, files up to 10 MB, the same per-IP daily limit as the free tools. 429 with Retry-After when you hit it.
- With Authorization: Bearer <REDUCTO_API_KEY>: your key, your plan's limits and credits, no page cap, files up to 50 MB.
- Only public http and https URLs on ports 80 and 443. Private, local and link-local addresses are blocked, including after redirects (up to 3).
- Supported: PDF, DOCX, PPTX, XLSX, PNG, JPG. Anything else returns 415.
- Errors come back as Markdown (or JSON with Accept: application/json) with a code in the X-Reducto-Error header.

## FAQ

### How is this different from r.jina.ai?

Same pattern, different input. Web readers turn HTML pages into Markdown; /read is for documents: PDFs (including scans), Word, PowerPoint and Excel files.

### Which model does it use?

Reducto Parse with r-1 mini, the fast Parse model. For scans with complex tables, call Parse with r-1 through the API.

### Do I need an account?

No, for the first 5 pages. For whole documents, create a free account (15,000 free credits) and send your API key as a Bearer token. Agents can sign up over HTTP: https://reducto.ai/auth.md.

### What does the comment at the top of the response mean?

It names the source URL and pages read, and on free reads it says how to get the rest of the document. It is an HTML comment, so Markdown renderers hide it.

### Can it read a URL behind a login?

No. /read fetches the URL from Reducto's servers without your cookies. Download the file and POST it to the API instead.

### Can I use it from an MCP client instead?

Yes: the hosted Reducto MCP server at https://mcp.reducto.ai/mcp has a parse tool that takes a URL or a local file.

### Can I keep tracked changes from a Word file?

/read returns the document text. To keep tracked changes and comments, use the Word to Markdown tool at /tools/docx-to-markdown.

## Related

- [Free document tools](https://reducto.ai/tools)
- [PDF to Markdown for LLMs and RAG](https://reducto.ai/tools/pdf-to-markdown)
- [Word to Markdown with tracked changes](https://reducto.ai/tools/docx-to-markdown)
- [Document MCP server for agents](https://reducto.ai/tools/reducto-mcp-claude-code-cursor)
- [r-1 mini announcement](https://reducto.ai/blog/reducto-r-1-mini-parse-model)

Last updated 2026-11-15
