A Jina Reader alternative with structure, not just markdown.
Reader-style tools convert a URL to markdown on request, and they do it well. Lyrenth serves the same clean markdown from a standing index of 1.1B+ pages, wrapped in a stable AIDocument: headings, links, images, JSON-LD, quality signals, cache truth, and the token economics of the read. Same convenience, more contract.
Where a reader proxy ends and an index begins:
| Reader-style proxies | Lyrenth | |
|---|---|---|
| Model | Convert on request, per caller | Standing index; one canonical document per URL |
| Output | Markdown text | AIDocument: markdown plus structure, signals, cache truth, economics |
| Schema | Text out | Published JSON Schema (draft-07), stable v2 envelope |
| Site owners | No relationship | Free dashboard, verification, canonical authorship, opt-out |
Honest scoping: for one-off conversions of pages nobody else reads, any good reader works. The index earns its keep when the same pages are read repeatedly, by you or by everyone.
Real pages, read through the production index, with the economics the API itself reports:
| Cloudflare Workers docs | 94,963 raw | 1,377 indexed | 98.5% saved |
| Kubernetes Pods concepts | 131,977 raw | 7,475 indexed | 94.3% saved |
| MDN: Overview of HTTP | 56,692 raw | 5,085 indexed | 91.0% saved |
Try it against the pages your agent already reads: npx -y lyrenth-mcp in any MCP client, or one GET /v1/read. Free tier is 2,000 reads a month, no card.