Summary
On August 6, 2026, AWS introduced Runtime instances for Amazon Bedrock AgentCore: persistent, managed EC2 infrastructure for production AI agents with multi-agent collaboration, GPU support, and sessions lasting up to 14 days. Multiple agents can share a host and session directory and collaborate without direct API calls between them.
What changed
Amazon Bedrock AgentCore gained Runtime instances, AWS-managed EC2 infrastructure where multiple agents deploy in a single runtime with their own dependencies, share a session directory, and collaborate without inter-agent API calls. It supports GPU acceleration, session stop/restart, containerized deployments, Python 3.11-3.14, and frameworks like CrewAI, LangGraph, LlamaIndex, and Strands, with sessions up to 14 days. AWS positions it as complementary to the existing microVM runtime (up to 8-hour invocations).
Why it matters
Long-running, stateful, multi-agent workloads need durable compute that microVMs and short invocations cannot provide. Persistent runtime instances make production multi-agent systems and GPU-heavy agent work practical on managed AWS infrastructure, deepening AgentCore as an agent operating platform.
Evidence excerpt
Runtime instances provide persistent, managed EC2 infrastructure for production agents with multi-agent collaboration, GPU support, and sessions up to 14 days; multiple agents share a host and session directory and collaborate without direct API calls. Positioned as complementary to microVMs (up to 8-hour invocations).