Voice Call MCP Server

🚀 Voice Call MCP Server

A Model Context Protocol (MCP) server that enables Claude and other AI assistants to initiate and manage voice calls using Twilio and OpenAI's (GPT-4o real-time model).

Use this as a foundation to quickly kickstart your exploration of AI-driven voice calls, save time, and develop more features on top of it.

🚀 Quick Start

This MCP server serves as a powerful tool for initiating and managing voice calls with the help of AI assistants. It integrates Twilio and OpenAI's GPT - 4o real - time model to provide a seamless voice - call experience.

✨ Features

• Initiate phone calls via Twilio 📞
• Process real - time voice using the GPT - 4o real - time model 🎙️
• Real - time language switching during calls 🌐
• Pre - built prompts for common call scenarios (e.g., restaurant reservations) ⍽️
• Automatic public URL tunneling with ngrok 🔄
• Securely handle credentials 🔒

📚 Documentation

Sequence Diagram

sequenceDiagram
    participant AI as AI Assistant (e.g., Claude)
    participant MCP as MCP Server
    participant Twilio as Twilio
    participant Phone as Destination Phone
    participant OpenAI as OpenAI
    
    AI->>MCP: 1) 发起外呼请求（POST /calls）
    MCP->>Twilio: 2) 通过Twilio API放置外呼
    Twilio->>Phone: 3) 拨打目标电话
    Twilio->>MCP: 4) 呼叫状态更新及音频回调（webhooks）
    MCP->>OpenAI: 5) 将实时音频转发到OpenaAI的实时模型
    OpenAI->>MCP: 6) 返回语音流
    MCP->>Twilio: 7) 发送语音流
    Twilio->>Phone: 8) 转发语音流
    Note over Phone: 双向对话继续进行<br>直到通话结束

Important Notes

⚠️ Important Note

1. Phone number format: All phone numbers must be in E.164 format (e.g., +11234567890).
2. Rate limits: Be aware of the rate limits and pricing of your Twilio and OpenAI accounts.
3. Voice conversations: The AI will process natural conversations in real - time.
4. Call duration: Be mindful of call durations as they affect OpenAI API and Twilio costs.
5. Public exposure: The ngrok tunnel exposes your server publicly (although protected by a random URL and random key).

Troubleshooting

Common error messages and solutions:

1. "Phone number must be in E.164 format"
   • Ensure the phone number starts with "+" and includes the country code.
2. "Invalid credentials"
   • Check TWILIO_ACCOUNT_SID and TWILIO_AUTH_TOKEN. You can copy them from the Twilio console.
3. "OpenAI API error"
   • Verify that OPENAI_API_KEY is correct and has sufficient credits.
4. "Ngrok tunnel failed to start"
   • Ensure NGROK_AUTHTOKEN is valid and not expired.
5. "OpenAI real - time fails to detect the end of voice input or has a delay."
   • Sometimes, there may be voice encoding issues between Twilio and the recipient's network operator. Try using a different recipient.

Contribution

💡 Usage Tip

Contributions are welcome! Here are areas we hope to improve:

• Implement support for multiple AI models (currently only GPT - 4o is supported).
• Add database integration to store local conversation history and make it accessible to the AI.
• Improve latency and response time to enhance the call experience.
• Enhance error handling and recovery mechanisms.
• Add more pre - built common scenario conversation templates.
• Implement improved call monitoring and analysis.

If you wish to contribute, please open an issue to discuss your ideas before submitting a pull request.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

Security

Do not include any sensitive information (such as phone numbers or API credentials) in GitHub issues or pull requests. This server handles sensitive communications; deploy responsibly and ensure all credentials are secure.