How to Connect MCP Servers to Cursor (Step-by-Step)

What Cursor MCP Support Looks Like

Cursor supports MCP servers natively through its Agent mode. When you add an MCP server, Cursor Agent can call those tools during conversations. Ask it to check your database, query an API, or pull metrics from a dashboard without leaving the editor.

MCP tools appear alongside Cursor's built-in tools (file editing, terminal, web search). The agent decides when to call them based on your request.

Step 1: Create Your mcp.json File

Cursor reads MCP configuration from .cursor/mcp.json in your project root. Create the file:

{
  "mcpServers": {
    "your-tool": {
      "url": "https://datafaucet.dev/api/mcp/YOUR_SERVER_ID/sse"
    }
  }
}

Replace YOUR_SERVER_ID with your DataFaucet server ID (visible on your dashboard after creating a server).

Step 2: Restart Cursor

After saving .cursor/mcp.json, restart Cursor or reload the window. The MCP connection initializes on startup.

To verify: open Cursor Settings > MCP. Your server should appear with a green status indicator. If it shows red, check the URL and ensure your server is deployed.

Step 3: Use Tools in Agent Mode

Switch to Agent mode (not Ask mode) in Cursor's chat panel. Agent mode is required for MCP tool calling.

Try a prompt that requires your tool:

• "Check the latest orders in our database"
• "What's the current deployment status?"
• "Pull the error rate from the last hour"

Cursor Agent will identify the relevant MCP tool, call it, and incorporate the result into its response.

Adding Multiple Servers

Add more servers to the same config file:

{
  "mcpServers": {
    "database": {
      "url": "https://datafaucet.dev/api/mcp/SERVER_1/sse"
    },
    "monitoring": {
      "url": "https://datafaucet.dev/api/mcp/SERVER_2/sse"
    },
    "project-management": {
      "url": "https://datafaucet.dev/api/mcp/SERVER_3/sse"
    }
  }
}

Each server name becomes the tool namespace. Cursor Agent sees all available tools and picks the right one based on context.

SSE vs Stdio Servers

Cursor supports two transport types:

• SSE (Server-Sent Events): Remote hosted servers. DataFaucet uses this. Just provide the URL.
• Stdio: Local processes that Cursor spawns. Requires a command and args.

DataFaucet servers are always SSE. No local process to manage, no Docker to run, no dependencies to install.

Troubleshooting

Server shows "disconnected" in settings:

Check that the URL is correct and the server is deployed. Visit the SSE URL in a browser. You should see an event stream connection.

Tools not appearing in Agent mode:

Make sure you're in Agent mode, not Ask mode. Ask mode doesn't call MCP tools.

"Tool call failed" errors:

Check your server's auth configuration on the DataFaucet dashboard. If the target API requires fresh credentials, you may need to re-record the server.

Creating a Server for Cursor

If you don't have an MCP server yet:

1. Go to datafaucet.dev
2. Enter any URL (your project management tool, database UI, monitoring dashboard)
3. Browse for 60 seconds. DataFaucet captures API calls.
4. Deploy. Copy the SSE URL into your .cursor/mcp.json.

The entire flow takes under two minutes. Your Cursor Agent gets typed tools with pre-configured auth and structured responses.

Team Setup

Commit .cursor/mcp.json to your repo so the whole team gets the same MCP tools. Each developer can also add personal servers in their global Cursor config at ~/.cursor/mcp.json.

Project-level config takes precedence for servers with the same name.