cursor s-0.46-windows-cmd-test

Cursor MCP servers extend the AI's capabilities by exposing external tools that can be invoked from conversations. This configuration demonstrates several commonly used MCP servers, including a sequential thinking server for step-by-step reasoning, Brave Search for web search capabilities, Puppeteer for browser-like actions, GitHub tools for interacting with repositories, and filesystem tools for local file interactions. To use them, add the desired server entries into your project or global mcp.json, then instruct the AI to call the appropriate server when a task requires those capabilities. When the AI selects a server, Cursor will spawn the specified tool, and terminal windows may appear to show the tool’s output in real time. You can combine multiple servers across a single session to perform complex workflows (e.g., search, summarize, and fetch GitHub data) efficiently.

Prerequisites:

• Windows 10/11
• Node.js v18 or higher
• Cursor IDE 0.46 or newer
• Basic command line knowledge

Step 1: Install Node.js and Cursor

• Download and install Node.js from https://nodejs.org/
• Install Cursor according to the official docs

Step 2: Create project-level MCP config (optional but recommended)

• In your project root, create a .cursor/mcp.json file with your MCP server definitions (examples shown below)

Step 3: Add MCP server definitions

• You can configure multiple MCP servers as shown in the mcp.json example. Each server uses npx to invoke the corresponding Model Context Protocol server package.

Step 4: Run Cursor and verify

• Launch Cursor and ensure the MCP servers appear green in the MCP panel when ready. If a server shows yellow, check the command and environment variables.

Code examples:

Project-level mcp.json

{
  "mcpServers": {
    "sequential_thinking": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-sequential-thinking"]
    },
    "brave_search": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-brave-search"],
      "env": {
        "BRAVE_API_KEY": "[your-key]"
      }
    },
    "puppeteer_browsing": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-puppeteer"]
    },
    "github_tools": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"]
    },
    "filesystem_tools": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-fs"]
    }
  }
}

Step 5: Start and test

• Open Cursor, load the project, and test each server by invoking its capabilities from the chat interface. Ensure you have any required API keys set as environment variables.

Tips:

• Only enable the MCP servers you need to minimize token usage and resource consumption.
• For sensitive keys (e.g., Brave API keys), prefer environment variables over hardcoding in scripts.
• If a project-level mcp.json doesn’t load on Windows 0.46 due to known issues, fall back to the global mcp.json or use the Cursor settings UI to add servers manually.
• Ensure Node.js version is at least v18; use nvm (or similar) to switch versions if needed.
• When testing browsers (Puppeteer) or websites, consider permission prompts and firewall rules that may block automated access.