How to Set Up MCP in Cursor: Complete Setup Guide in 2026

Why Use MCP with Cursor?

Cursor already brings powerful AI assistance directly into your code editor — autocompletes, chat, and refactoring all powered by large language models. But by itself, Cursor operates in a sandbox. It knows your open files. It does not know your GitHub repos, your running databases, or your production logs.

MCP bridges that gap. When you connect an MCP server to Cursor, your AI assistant gains real-time access to external tools and data. You can ask Cursor to:

• Pull the latest commits and issues from your GitHub repository
• Query your PostgreSQL or MySQL database directly from a chat prompt
• Search the web without leaving the editor
• Read and write files outside the current project directory
• Interact with Slack, Notion, or other workplace tools

The result is an AI coding assistant that is genuinely aware of your entire development context — not just the file you happen to have open.

Prerequisites

Before starting, make sure you have the following:

How to Enable MCP in Cursor

Cursor does not expose MCP settings through a dedicated UI panel yet. Instead, MCP configuration is managed through a JSON file that Cursor reads on startup. Here is the full walkthrough.

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem"],
      "env": {
        "allowedDirectory": "/Users/yourname/projects"
      }
    }
  }
}

Adding MCP Servers to Cursor — JSON Config Deep Dive

The JSON configuration format follows the standard MCP client specification. Each server entry requires at minimum a command and args field. Here is the anatomy of a full Cursor MCP config:

{
  "mcpServers": {
    "server-name": {
      "command": "npx",           // or "node" for local JS servers
      "args": ["-y", "package-name"],
      "env": {
        "KEY": "VALUE"            // server-specific environment variables
      },
      "scope": "project"          // "project" or "global"
    }
  }
}

The scope field determines whether the server is available to just the current project or to all Cursor projects. Use project scope for team-specific servers and global for universal tools like web search.

Testing MCP Connections Inside Cursor

After saving your configuration file, restart Cursor. You should see an MCP indicator (usually a small tool icon) in the Cursor AI chat panel when servers are active.

To verify a connection is working, try asking Cursor a question that triggers the MCP server. For example, with the filesystem server:

If Cursor responds using the MCP tool, the connection is live. If it fails or says the tool is unavailable, check the console for error messages and verify your JSON config is valid.

You can also use the MCP Inspector CLI to test server connections independently before connecting them to Cursor.

Recommended MCP Servers for Cursor Users

Here are the most useful MCP servers to connect to Cursor in 2026, all available on the MCPize Marketplace:

Common Issues and Fixes

Boost Your Cursor Workflow with Raycast

While Cursor handles your AI-assisted coding, Raycast supercharges the rest of your macOS workflow. Think of it as a universal command bar that launches apps, runs scripts, manages clipboard history, and integrates with dozens of tools — all without leaving the keyboard.

Used together, Cursor and Raycast form a powerful pair: Cursor for AI-first code editing, Raycast for everything else. Many developers find that once they go all-in on Raycast, switching back to Spotlight feels like downgrading.

Get Raycast free →

Get Started Today

Setting up MCP in Cursor takes less than 15 minutes. Here is your action plan to get running right now:

1. Open Cursor Settings (Cmd + ,) and navigate to Features → MCP. Click Edit JSON.
2. Add the filesystem MCP server using the JSON config example above. Replace /Users/yourname/projects with your actual project directory.
3. Restart Cursor and look for the MCP tool indicator in the AI chat panel.
4. Ask Cursor to read or write a file outside your current project to verify the connection works.
5. Browse the MCPize Marketplace and add more servers — the GitHub and Brave Search servers are particularly powerful additions.

Once you have your first MCP server connected and working, you will immediately see how much more context-aware your Cursor AI assistant becomes. From there, it is a matter of exploring which servers fit your workflow best.