AWS published a technical walkthrough for building a telephony AI host that answers a phone number and takes a restaurant order, guiding the caller from greeting to confirmation. The system uses Amazon Bedrock AgentCore to host and run the agent and Amazon Nova 2 Sonic to handle real-time speech, with the agent connected to a restaurant backend through the Model Context Protocol (MCP).
The post covers deploying the full stack with the AWS Cloud Development Kit (CDK) and bridging an incoming phone call into the agent using a Session Initiation Protocol (SIP) gateway running on Amazon Elastic Container Service (ECS) and AWS Fargate. It also describes warming the agent session while the phone is still ringing, so callers do not encounter silence when the call connects.
Who should care
Developers and teams working on voice agents or telephony integrations on AWS can use the walkthrough as a reference architecture for connecting speech models and backend systems to a live phone line.