4 agent tools live

Let your AI agent use the index directly.

MCP Index implements WebMCP — the emerging web standard (document.modelContext) that lets a page expose tools to an in-browser AI agent. Instead of describing servers and making you copy commands, the agent calls the index itself: it searches the curated set and installs a server for you, live on this page.

The tools it exposes

Four tools, scoped to trusted listings only, so an agent acting on your behalf can never be steered to a flagged or dead server.

search_index

Explore trusted MCP servers by keyword or category. Flagged, spam, delisted, and dead entries are excluded — the agent only ever sees the curated set.

get_server

Pull the full record for one server: what it does, our editorial take, repo and package links, verification status, and every install option.

install_server

Hand the user the exact install command for a server, list any required credentials, and open its page for review. The must-have: explore, then install, without leaving the chat.

list_categories

List every category and how many live servers it holds, so the agent can narrow a search precisely.

How to connect

01An agent that speaks WebMCP

Use a browser with the native Web Model Context API (Chromium, behind the WebMCP origin trial) or the MCP-B extension, which bridges this page's tools to Claude, ChatGPT, Gemini, Cursor, or Copilot. No account or API key on our side.

02Open this site and ask

With the page open, the four tools above register automatically. Ask your agent to find or install an MCP server — it does the rest.

Try asking

Find me a trusted MCP server for talking to Postgres and install it.
What web-scraping MCP servers does the index have? Which is hand-reviewed?
Install the top GitHub MCP server and tell me what credentials I need.

Trusted set only

The agent tools never surface a listing we've flagged as bad-faith, a delisted placeholder, or a server with a dead repo. Every result carries its verification status — hand-reviewed, or auto-verified by the nightly checks — so the agent can tell you exactly what it's about to install.

How verification and flagging work →

Browse the index yourself →