Voice Call MCP Server is an MCP server that enables AI assistants to initiate and manage two-way telephone calls using Twilio and OpenAI's GPT-4o Realtime model. The server connects Model Context Protocol clients like Claude Desktop to telephony infrastructure via Twilio, using ngrok to expose necessary webhooks for live bidirectional audio streaming. Developers, automated agents, and business operations teams use this tool to delegate voice-based tasks directly to an AI model. Once triggered by a user prompt, the server places outbound calls to destination numbers formatted in standard E.164 notation. Audio is routed to OpenAI for immediate real-time processing and voice generation, while the response is streamed back through Twilio to the recipient on the line. The implementation supports multilingual dialogues, dynamic mid-call language switching, and call recording configurations. By handling telephony webhooks and audio bridging locally, Voice Call MCP Server allows assistants to handle practical communication workflows like appointment bookings, reservation confirmations, and status updates directly over standard cellular and landline phone networks.
Category: Communication & Messaging
Tags: calling, telephony, voice, voip
bash git clone https://github.com/lukaskai/voice-call-mcp-server.git cd voice-call-mcp-server 3. Install dependencies and build the server: bash npm install npm run build 4. Add the server entry to your Claude Desktop configuration file (claude_desktop_config.json): json { "mcpServers": { "voice-call": { "command": "node", "args": ["/path/to/your/voice-call-mcp-server/dist/start-all.cjs"], "env": { "TWILIO_ACCOUNT_SID": "your_account_sid", "TWILIO_AUTH_TOKEN": "your_auth_token", "TWILIO_NUMBER": "your_e.164_format_number", "OPENAI_API_KEY": "your_openai_api_key", "NGROK_AUTHTOKEN": "your_ngrok_authtoken" } } } } 5. Restart Claude Desktop to enable the server.Part of MCP Servers
Voice Call MCP Server allows AI clients such as Claude Desktop to place real outbound phone calls. It uses Twilio to dial phone numbers and connects the live audio stream directly to OpenAI's GPT-4o Realtime model. This allows the AI to converse naturally with human recipients, handle scheduling or inquiries, switch languages during calls, and optionally record conversations.
Voice Call MCP Server works with any client supporting the Model Context Protocol that can manage stdio-based local server connections. Claude Desktop for macOS and Windows is the primary supported client tested in the project repository. It can also be configured with other MCP-compatible development tools and custom agents capable of launching Node.js processes.
To install the server, verify Node.js 22 is installed, clone the GitHub repository, and run npm install followed by npm run build. Then, add the server command pointing to dist/start-all.cjs in your Claude Desktop configuration file alongside required credentials including your Twilio Account SID, Twilio Auth Token, Twilio Number, OpenAI API key, and Ngrok Authtoken.
Yes, Voice Call MCP Server is open-source software licensed under the MIT License. The code is publicly hosted on GitHub, allowing developers to inspect, modify, and extend the implementation for their own voice calling agents, internal workflows, or customer communication tooling without restrictive commercial licensing barriers.
Voice Call MCP Server requires all phone numbers to be formatted according to the international E.164 standard. This standard mandates that phone numbers must start with a plus sign followed by the country code, area code, and subscriber number without any spaces, dashes, or punctuation, such as +11234567890.