I'm happy to walk you through how the OpenAI realtime model interacts with MCP servers via the LiveKit agent framework. Here’s a simple breakdown: • MCP Server Initialization: The MCP servers are set up and passed to the agent during its initialization. The agent then connects to these servers and gathers the available tools. • Tool Integration: The tools from the MCP servers are transformed into a format that the agent can work with, making them accessible to the OpenAI model. • Execution Flow: When the model chooses to use a tool, the agent takes care of communicating with the MCP server and brings back the results to the model. This way, the OpenAI model can use the tools smoothly without needing to interact directly with the MCP servers. If you have any questions or want me to explain anything further, just let me know—I’m here to help! Sources: Tool definition and use | LiveKit Docs | Tool definition and use | LiveKit Docs | livekit-agents/livekit/agents/voice/agent.py | plugins/openai/src/realtime/api_proto.ts