HomeAI NewsAmazon Bedrock AgentCore Runtime Instances host multi-day agent workflows

Amazon Bedrock AgentCore Runtime Instances host multi-day agent workflows

AWS introduced Runtime Instances for agent hosting with GPUs, persistent storage, and sessions lasting up to 14 days.

AWS introduced Amazon Bedrock AgentCore Runtime Instances for hosting agents that run for days, not hours. Runtime Instances use AWS managed EC2 infrastructure. They add GPUs, persistent Amazon EBS volumes, and multi-day sessions that last up to 14 days. A single instance can host multiple agents that share a filesystem.

Amazon Bedrock AgentCore offers two compute options. MicroVMs provide serverless sessions with fast cold starts and isolation. Runtime Instances provide AWS managed EC2 capacity for persistent workflows. Both options support custom frameworks such as CrewAI, LangGraph, LlamaIndex, and Strands Agents, plus MCP and A2A integration.

Builders can colocate multiple agents on one EC2 instance by sharing a capacity provider and runtimeSessionId. Those agents then share a filesystem and build on each other’s output. Operators can use AWS Savings Plans and On-Demand Capacity Reservations for Runtime Instances. Enterprise teams gain session persistence for workflows that span multiple days.

The example pipeline uses three agents: Compose, Deliver, and Screen. One agent runs a generative audio model on the instance GPU. The other two agents open the .wav file from a shared volume. The post covers capacity providers, artifact deployment, agent collaboration, and multi-day persistence.

What matters

  • AWS now offers Runtime Instances for Amazon Bedrock AgentCore agents that need multi-day sessions.
  • Builders gain shared sessions, persistent EBS storage, and GPU access across multiple colocated agents.
  • Watch how capacity providers manage scaling as multiple agents share one runtimeSessionId.

Why it matters

Watch how capacity providers manage scaling as multiple agents share one runtimeSessionId.

This GenAI News article was prepared in original wording using reporting and materials published by AWS Machine Learning Blog. Source reference: https://aws.amazon.com/blogs/machine-learning/build-a-multi-agent-music-production-pipeline-on-amazon-bedrock-agentcore-runtime-instances/.

Drafted by the GenAI News review pipeline.

latest articles

explore more