# AgentLens

> AgentLens grades how ready a website is for AI agents. The free scan is a
> static pre-check — evidence about a site's agent-facing surface, scored in
> seconds. The paid audit (https://agent-chi-lime.vercel.app/audit) measures behaviour: repeated
> real Claude Code and cursor-agent runs with confidence intervals.

## Using AgentLens

- [Scan a domain](https://agent-chi-lime.vercel.app/): paste a domain, get a scored report in ~10 seconds
- [Leaderboard](https://agent-chi-lime.vercel.app/leaderboard): domains ranked by agent-readiness, per category
- [Documentation](https://agent-chi-lime.vercel.app/docs): what the checks measure and how scoring works
- [OpenAPI](https://agent-chi-lime.vercel.app/api/openapi.json): the full HTTP API contract

## API quick reference

- POST https://agent-chi-lime.vercel.app/api/scan with {"url": "https://example.com"} — run a pre-check
- GET https://agent-chi-lime.vercel.app/api/score/{domain} — the stored result; a 404 tells you how to trigger the scan
- GET https://agent-chi-lime.vercel.app/api/badge/{domain} — an SVG badge with the scan score and grade

## For agents

- MCP server (no auth): `claude mcp add --transport http agentlens https://agent-chi-lime.vercel.app/api/mcp`
- [skill.md](https://agent-chi-lime.vercel.app/skill.md): the fix-loop playbook — tunnel localhost, scan, fix worst-first, rescan
- [agents.md](https://agent-chi-lime.vercel.app/agents.md): what you may do here, rate limits, error envelope
- Errors are JSON with a stable `code` and a `next_action` you can follow mechanically.
- Rate limit: 10 scans/min/IP. No auth needed for scanning.