AWS Releases Modular Voice Travel Concierge Template with Bedrock AgentCore, Knowledge Bases, and Nova Sonic
Amazon Web Services has published a developer-ready template for a voice-enabled travel concierge, combining Bedrock AgentCore, real-time speech with Nova Sonic, and retrieval-augmented airline policy answers using managed Knowledge Bases. The architecture separates frontend, agent, and backend for scalable, modular development, and deploys with a single AWS CDK script.
What changed?
Amazon Web Services has released a reference implementation and step-by-step guide for building a voice travel concierge using three Bedrock services: AgentCore (for agent orchestration and secure tool invocation), Nova Sonic (for real-time speech-to-speech in the browser), and managed Bedrock Knowledge Bases (for retrieval-augmented, policy-grounded answers). The template demonstrates how everyday developers can enable travelers to check flights, change seats, update meal preferences, or ask policy questions by voice within an airline’s web application. The architecture strictly separates the frontend, agent, and backend services for modular code and independent scaling. Integration with backend airline systems is standardized via Model Context Protocol (MCP) through AgentCore Gateway, exposing REST APIs and Bedrock Knowledge Bases as discoverable tools. Deployment is automated for AWS via a single Cloud Development Kit (CDK) script.
Why does it matter to an everyday developer?
This template greatly reduces the complexity of adding conversational, policy-aware voice features to any travel or logistics web experience. Developers do not need to manage transcription, agentic context, backend orchestration, or policy retrieval manually. The separation of layers allows frontend, voice agent, and backend teams to iterate separately. The MCP protocol enables backend APIs and knowledge bases to be attached or updated as tools without changing agent or frontend code. Bedrock Knowledge Bases require no custom retrieval, embedding, or Lambda boilerplate, and documentation syncs to S3 in near real time. The deployment process is streamlined: the developer can use AWS CDK for one-shot deployment and focus on feature adaptation, not infrastructure.