Deploy voice agents with Pipecat and Amazon Bedrock AgentCore Runtime – Part 1

2026-03-25 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Software Development & Engineering · Depth: Intermediate, long

Summary

This post, a collaboration between AWS and Pipecat, details deploying real-time voice agents using Pipecat on Amazon Bedrock AgentCore Runtime. It addresses challenges like low-latency streaming, security isolation, and dynamic scalability for natural, human-like conversations. AgentCore Runtime provides a secure, serverless environment with isolated microVMs, auto-scaling, and cost-effective resource utilization for sessions up to 8 hours. The article explores four network transport approaches for client-to-agent connections: WebSockets, WebRTC (TURN-assisted), managed WebRTC, and telephony integration. It provides practical deployment guidance and code samples for each, including configuring AgentCore Runtime with VPC for WebRTC and using AWS-native TURN with Amazon Kinesis Video Streams (KVS).

Key takeaway

For AI Engineers building real-time voice agents, selecting the appropriate network transport is crucial for user experience and system reliability. You should start with WebSockets for initial prototyping due to its simplicity, then transition to WebRTC with AgentCore Runtime on VPC mode or a managed provider for production deployments to ensure low latency and resilience. If your agents require traditional phone call integration, explore telephony provider integrations to maintain conversational flow.

Key insights

Achieving natural, low-latency voice AI requires robust streaming architectures and scalable runtime environments.

Principles

• Low latency is critical for natural voice agent interactions.
• Security and scalability are paramount for production voice agents.
• Bidirectional streaming is essential for real-time audio processing.

Method

Deploy Pipecat voice agents as ARM64 containers on Amazon Bedrock AgentCore Runtime, configuring network transport via WebSockets, WebRTC (TURN-assisted, managed, or KVS-integrated), or telephony for optimal latency and reliability.

In practice

• Use WebSockets for rapid prototyping of voice agents.
• Implement WebRTC with TURN servers for production-grade latency.
• Configure AgentCore Runtime with VPC for WebRTC UDP transport.

Topics

• Voice Agents
• Amazon Bedrock AgentCore Runtime
• Real-time AI
• WebRTC
• Streaming Architectures

Code references

• pipecat-ai/pipecat
• pipecat-ai/pipecat-examples
• coturn/coturn
• aws-samples/sample-voice-agent

Best for: AI Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• Scalable voice agent design with Amazon Nova Sonic: multi-agent, tools, and session segmentation — Multi-agent architectures and session segmentation are key to building scalable, low-latency voice agents.
• Voice Live API now supports WebRTC (Preview) — WebRTC integration in Voice Live API enables low-latency, real-time voice interactions for AI agents.
• AWS Bedrock Agents: The Beginner’s Guide — AWS Bedrock Agents provide a managed platform for building and deploying AI agents with integrated orchestration and security.
• Public preview: Voice-native agents in Microsoft Foundry — Microsoft's new Foundry Agent Service and Voice Live API simplify building real-time, speech-to-speech AI agents on Azure.
• Introducing stateful MCP client capabilities on Amazon Bedrock AgentCore Runtime — Stateful MCP on Amazon Bedrock AgentCore Runtime enables interactive, multi-turn AI agent workflows.
• How OpenAI delivers low-latency voice AI at scale — OpenAI's WebRTC rearchitecture uses a split relay-transceiver model for scalable, low-latency voice AI.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.