🧩 How to Connect MCP Server to Claude, ChatGPT, Cursor, Windsurf & Perplexity

---

Claude Desktop

1. Install prerequisites
   • Install Claude Desktop for macOS or Windows.
   • Install Node.js LTS and verify installation:

node --version

1. Locate the configuration file
   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
2. Add your MCP configuration Replace the content with:

{
  "mcpServers": {
    "MCP Server": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.mcpfy.ai/<token>/stream",
        "--header",
        "Authorization: Bearer {{value}}"
      ]
    }
  }
}

3. Replace {{value}} with your actual Bearer token.
4. Restart Claude Desktop
   • Quit and reopen Claude.
   • Look for the MCP indicator near the message box to confirm it loaded successfully.
5. Test it Try asking:
   • “Use my MCP Server to list connected tools.”
   • “Fetch user data via my MCP Server.”

ChatGPT

1. Enable Developer Mode
   • Go to Settings → Apps & Connectors → Advanced Settings.
   • Toggle Developer Mode ON.
2. Add your MCP connector
   • Navigate to Settings → Connectors → Create.
   • Enter:
     • Name: MCP Server
     • Description: Connects to your MCP endpoint
     • Connector URL:

https://mcp.mcpfy.ai/<token>/mcp

• Add authentication (e.g., Bearer token) if required.
  

2. Use it in chat
   • Start a new chat → click + → More → MCP Server.
   • Ask naturally, e.g. “Fetch open tasks from my MCP Server.”

💡 Note: ChatGPT supports only remote HTTPS MCP servers. To use local servers, expose them with ngrok or Cloudflare Tunnel.

---

💻 Connect MCP Server in 

Cursor

1. Open MCP settings
   • Go to Settings → Developer → MCP Servers.
   • Or edit ~/.cursor/mcp_config.json directly.
2. Add configuration Paste this JSON:

{
  "mcpServers": {
    "MCP Server": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.mcpfy.ai/<token>/stream",
        "--header",
        "Authorization: Bearer {{value}}"
      ]
    }
  }
}

2. Test it In Cursor’s chat, try: “Use my MCP Server to summarize files in this repo.”

---

🪶 Connect MCP Server in 

Windsurf

1. Open MCP Servers settings
   • In Windsurf, go to Settings → Cascade → MCP Servers.
   • Click Add Custom Server.
2. Enter configuration

{
  "mcpServers": {
    "MCP Server": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.mcpfy.ai/<token>/stream",
        "--header",
        "Authorization: Bearer {{value}}"
      ]
    }
  }
}

2. Save and restart Windsurf Restart the IDE to apply the configuration.
3. Enable the MCP Server
   • Go back to the MCP list and toggle your MCP Server ON.
   • Test with: “Fetch API results using my MCP Server.”

---

Perplexity AI

1. Install the helper app (for local MCPs)
   • macOS: Install PerplexityXPC if prompted.
   • Windows/Linux: Use remote MCPs directly (no helper needed).
2. Add the MCP Connector
   • Go to Settings → Connectors → Add Connector → MCP Connector.
   • Enter:
     • Name: MCP Server
     • Command:

npx -y mcp-remote https://mcp.mcpfy.ai/<token>/stream --header "Authorization: Bearer {{value}}"

2. Save and enable
   • Wait for it to show Running.
   • Enable it in your Sources list.
3. Try it Ask: “Use my MCP Server to find top AI papers this week.”

---

🧰 Troubleshooting Tips

If your MCP Server doesn’t show up:

• ✅ Verify Node.js and npx are installed globally.
• ✅ Replace {{value}} with your real Bearer token.
• ✅ Ensure the server endpoint is correct:

https://mcp.mcpfy.ai/<token>/stream

• ✅ Run it manually for debugging:

npx -y mcp-remote https://mcp.mcpfy.ai/<token>/stream

• ✅ Check logs:
  • Claude: ~/Library/Logs/Claude/mcp*.log
  • Cursor/Windsurf: ~/.codeium/… or terminal output
  • Perplexity: Connector logs panel

---

🚀 Conclusion

That’s it! You’ve connected your MCP Server to:

• Claude
• ChatGPT
• Cursor
• Windsurf
• Perplexity

Now your AI tools can securely execute actions, access external APIs, and interact with your data in real-time, all powered by your single MCP endpoint.