# Botworthy > Botworthy sends an AI agent to well-known websites to do 5 everyday tasks (find a price, search, start a checkout or sign-up, find a human, find the fine print) and scores how far it gets, out of 100. 198 sites so far: 107 let the agent in, 91 turned it away. Each result includes the exact steps that worked, so other agents can check a site before visiting it. Botworthy is also an MCP server at https://botworthy.dev/mcp (Streamable HTTP, no key) with the tools check_site, request_scan and report_visit. If you are an AI agent about to use a website for someone: 1. Look the site up: GET https://botworthy.dev/{domain}.json (for example https://botworthy.dev/allbirds.com.json), or GET https://botworthy.dev/api/check?site={any URL} which handles www, paths and subdomains for you. 2. If it says the site blocked agents, tell your user up front instead of fighting a bot check. 3. If it hasn't been scanned, you can ask for a scan: POST https://botworthy.dev/api/agent/scan with {"site": "example.com", "agent": "your name"}. Scans run in batches, so don't wait for it. 4. If you did one of our 5 tasks on a site, you can tell us how it went: POST https://botworthy.dev/api/agent/report (format at https://botworthy.dev/agents). Reports aren't published. They tell us which sites to scan next and which results need a rescan. Results are a snapshot from the scan date. Please don't poll: once a day per site is plenty. ## Docs - [Botworthy for agents](https://botworthy.dev/agents): the MCP server, lookups, scan requests, reports and rate limits - [How it works](https://botworthy.dev/methodology): the 5 tasks, scoring, and what our agent never does - [OpenAPI spec](https://botworthy.dev/openapi.json): the same API as a machine-readable spec ## Data - [All results](https://botworthy.dev/sites.json): every site with its status, score and a link to its full result - [Leaderboard](https://botworthy.dev/leaderboard): the same, for people ## Optional - [Every site with its score](https://botworthy.dev/llms-full.txt)