Voice Call MCP Server

Example Use Cases

• 1

  Place outbound phone calls and control call flow programmatically via an MCP endpoint.

• 2

  Process and respond to real-time call audio using OpenAI's GPT-4o Realtime model during two-way conversations.

• 3

  Make language-specific calls or use pre-built conversation templates (e.g., reservations, appointment changes) with automatic public webhook exposure via ngrok.

Description

This server enables AI assistants (e.g., Claude) to place outbound phone calls via Twilio, stream call audio to OpenAI's GPT-4o Realtime model, and relay AI-generated audio back to the call in real time. It includes features like runtime language switching, pre-built conversation prompts, automatic ngrok tunneling for webhooks, and secure environment-based credential handling. The project is intended as a developer-friendly base to build AI-powered voice calling workflows and integrations with Claude Desktop. It reduces boilerplate for realtime audio routing and lets developers focus on conversation logic and UX.