Add AgentReady as an MCP server so your agent can discover, query, and plan against public websites. Sites are indexed automatically when a workflow needs one. One connection gives you access to every site in the directory.
If you're using a browser-based AI agent on Chrome 149 or later, AgentReady's tools are already available — no URL to paste, no npm install, no config file. The tools register automatically when you visit this page.
Tools registered via WebMCP
list_sites — list all indexed websites
get_site_capabilities — index on demand, then inspect a site's manifest and endpoints
ask_site — ask a cited question about any site
plan_site_action — index on demand, then create a grounded, read-only plan
submit_site — index a new website
refresh_site — refresh content, or index a new site on demand
rate_answer — send quality feedback on an answer
Not on Chrome 149+? Use the MCP URL below or the local stdio bridge instead.
Paste this URL into any MCP-compatible client that supports HTTP transport.
https://www.agentready.it.com/api/mcpOpen your MCP client config
For clients that use an mcpServers configuration, open the client's config file.
Add the "agentready" server
{
"mcpServers": {
"agentready": {
"command": "npx",
"args": ["-y", "@agentreadyweb/mcp"]
}
}
}If you already have other MCP servers, add agentready inside the existing mcpServers object.
Restart your MCP client
Reopen the client. You'll see seven tools covering discovery, capabilities, questions, plans, indexing, freshness, and feedback.
Try it
Ask your agent:
"Use list_sites, inspect the site's capabilities, then create a grounded plan for my request."
Grade any site's agent-readiness
npx @agentreadyweb/mcp grade yoursite.com
Checks llms.txt, sitemap, robots.txt, JS-rendering, crawlability, and index status. Exits non-zero below a B — drop it into CI to gate deploys.
Ask any site a question
npx @agentreadyweb/mcp ask stripe.com "what is the test card number?"
Cited answer in seconds. Auto-indexes sites that aren't in the index yet (~60s).
Index a new site
npx @agentreadyweb/mcp index https://yoursite.com
Crawl a new public website so agents can query it with cited answers. JS-rendered sites are supported.
Keep your index fresh from CI
# GitHub Actions — after your deploy step - run: npx @agentreadyweb/mcp refresh yoursite.com
Re-crawls after every deploy so agents always see current content.
Open Cursor Settings → MCP
Go to Cursor Settings → Features → MCP Servers → Add new MCP server.
Add the server URL
Set type to HTTP and paste:
https://www.agentready.it.com/api/mcp
Don't need MCP? Query any indexed site directly over HTTP.
Find your site ID
Submit a URL on the home page — the result URL contains your site ID.
POST to /ask
curl -X POST https://www.agentready.it.com/api/sites/{siteId}/ask \
-H "Content-Type: application/json" \
-d '{"query": "What does this site do?"}'Response:
{
"answer": "...",
"sources": [{ "url": "...", "title": "..." }]
}list_sites()Returns all websites indexed by AgentReady with domain, title, and summary.
get_site_capabilities(domain)Automatically indexes a public site when needed, then returns its capability manifest, freshness state, schemas, and available HTTP/MCP endpoints.
plan_site_action(domain, request)Automatically indexes a public site when needed, then creates a grounded, read-only plan with evidence, risks, confirmation requirements, and a durable receipt.
submit_site(url)Index any website so it can be queried. Takes ~60 seconds. Handles static HTML, Next.js SPAs, and sites with llms.txt.
ask_site(domain, query)Ask a question about any public site. AgentReady indexes it automatically if needed, then returns a cited answer grounded in its actual content.
Example: ask_site("clueso.io", "What are the pricing plans?")
refresh_site(domain)Re-crawl a site for new or changed content. If it has never been indexed, this performs the initial indexing automatically. Takes ~60 seconds.
rate_answer(domain, rating)Send 1–5 quality feedback, optionally tied to the exact ask_site request.
Plans are read-only today. Permissioned customer actions require explicit confirmation and an enabled connector; they are not exposed as an unrestricted MCP execute tool.
Want this for your org or product?
If you want AgentReady deployed privately — for your internal sites, your product's knowledge base, or your team's tooling — get in touch.