# Rabbytte

> A small, human-curated index of the good web. Every site was submitted and approved by a person — no crawling, no SEO sludge. The whole index is readable by machines through a free, no-auth JSON API.

Rabbytte exists to give AI systems high-signal, human-vetted grounding data. Sites carry titles, descriptions, tags, sitelinks, street addresses with coordinates when published, and preview images.

## API
- [Search](https://rabbytte.com/api/search?q=example): GET /api/search?q=&page=&per= — full-text search with rank scores
- [Ask](https://rabbytte.com/api/ask?q=example): GET /api/ask?q= — a question answered strictly from the index, with sources (rate-limited)
- [Images](https://rabbytte.com/api/images?q=example): GET /api/images?q= — matching sites' stored preview images
- [OpenAPI spec](https://rabbytte.com/openapi.json)
- [Human docs](https://rabbytte.com/api-docs)

## MCP
- npm package `rabbytte-mcp` plugs the index into Claude, Cursor, and any MCP client.

## For humans
- [Home](https://rabbytte.com/)
- [Submit a site](https://rabbytte.com/submit) — humans only; a person reviews everything