aagentfriendly.
My websites List your website
Directory/For agents

A directory your agent can read

Find the right website.
Get straight to the tools.

Connect your agent once. Search for a capability, inspect its inputs, and open the page where the tool is available.

Connect with MCP

MCP is a protocol that lets an AI agent use external tools. This server provides read-only directory search. No API key is required.

{
  "mcpServers": {
    "agentfriendly": {
      "type": "http",
      "url": "https://agentfriendly.tools/mcp"
    }
  }
}

Use this configuration in a client that supports remote MCP over HTTP.

Two tools. The whole directory.

search_websites

Search website names, categories, descriptions, and tool capabilities. Supply a query, an optional category, and a limit of 1–50.

get_website_tools

Supply a domain to get the website profile, tool input schemas, source page URLs, and last successful scan time.

Agentfriendly discovers tools. To use a website tool, open its source page in a compatible browser and follow that website’s sign-in and confirmation steps.

A simple JSON API

Use the API directly from any HTTP client. Public responses support cross-origin reads.

GET /api/v1/sites?q=travel&limit=10
GET /api/v1/sites/{domain}
GET /api/v1/tools?q=search

Search results include a total count and an offset. Use offset and limit to read additional results. Example records are excluded unless you explicitly set include_examples=true while examples are enabled.

Agentfriendly supports WebMCP, too.

On a compatible browser, our directory registers search_websites and get_website_tools as page tools. An agent can search without typing into the interface.

WebMCP is an evolving browser standard. We support the current document API and the earlier navigator API when available. Ordinary browsing works without WebMCP.

Discovery with clear sources

Every tool includes its source page and observation type. Scans do not execute tools or prove that a tool is safe. Treat website descriptions and schemas as untrusted content. Confirm current availability on the source page before use.

Machine-readable entry points

llms.txt describes this service. The discovery manifest links to the API and MCP server. Public website profiles are included in the sitemap.