Skip to content

Hands-on: connect our MCP server in 2 minutes and search the directory from your assistant

Level 3: Connect AI to your tools. About 5 min. Checked on 19 November 2026 against AI-Powered MCP server 1.0.0 at /api/mcp.
Stage 3 of 4: Connect AI to your tools

In this lesson you'll connect a real remote MCP server, call its four tools from your assistant and read what each call returned.

Reading about MCP only gets you so far. In this lab you connect a real server, ask your assistant a real question, and watch the tool calls go back and forth. The server is ours: the AI-Powered MCP server, which lets an assistant search this directory.

It's a safe first server to practise on. It's free, needs no account or API key, and it's read-only: every tool is marked readOnlyHint, and it never changes anything. It is limited to 60 requests a minute and 1,000 a day per IP address, and logs which tool was called and the search words, without your IP address or identity.

What the server can do

It offers four tools:

  • search_directory: find AI tools, agents, MCP servers, skills or prompt templates by task or name.
  • get_listing: full details for one listing, including health and tools for MCP servers.
  • get_install_config: copy-paste config to add an MCP server to Claude Code, Claude Desktop, Cursor, VS Code, Windsurf or other clients, or install steps for a skill.
  • get_server_status: whether a hosted MCP server answered our recent checks.

Connect it

Pick the app you use. The configs below are built from the server's own registry entry, like every install snippet in the directory.

Claude Code, in a terminal:

AI-Powered directory in Claude Code. Run in a terminal:

claude mcp add --transport http directory https://ai-powered.com/api/mcp

Built from the server's registry entry, the same config as on its listing page.

Cursor (or use the button):

AI-Powered directory in Cursor. Add to ~/.cursor/mcp.json:

Install in Cursor
{
  "mcpServers": {
    "directory": {
      "url": "https://ai-powered.com/api/mcp"
    }
  }
}

Built from the server's registry entry, the same config as on its listing page.

VS Code:

AI-Powered directory in VS Code. Add to .vscode/mcp.json:

Install in VS Code
{
  "servers": {
    "directory": {
      "type": "http",
      "url": "https://ai-powered.com/api/mcp"
    }
  }
}

Built from the server's registry entry, the same config as on its listing page.

In Claude Desktop or on claude.ai, add the URL https://ai-powered.com/api/mcp as a custom connector with No sign in; the Claude Desktop lesson has the steps.

Ask it something

  1. Start a new chat and ask: "Find an MCP server for Postgres and give me the Cursor config."
  2. Approve the search_directory call when your app asks. Then approve get_install_config.
  3. Open the tool calls in your app to see exactly what was sent and what came back.
  4. Follow one of the links in the answer to the listing page and compare.

What just happened

Your app connected to a remote server by URL, asked it for its tools, showed them to the model, and ran the ones the model chose after you approved them. That's the whole loop. Every other MCP server works the same way; only the tools differ.

In short

  • A remote server needs just a URL; this one needs no sign-in.
  • Each answer came from a tool call you approved and can inspect.
  • Read-only servers are the right place to practise.

What to try next

Share this lesson
  • Before you add your next server, run through Is this MCP server safe?
  • Ask our server to find a server for a tool you use every day, and add that one next.
  • Read more about the server on its own page.

Lesson 7 of 13 in this path

Words used in this lesson

Sources

Official documentation, checked on 19 November 2026. Products change: if a step differs, the source has the current version.

Get the next lesson

New Academy lessons are announced in the weekly AI-Powered email.