AWS introduced AgentCore runtime instances enabling deployment of AI agents on user-specified Amazon EC2 instance types without manual infrastructure management. This approach complements AgentCore’s existing microVM-based serverless runtime, accommodating long-running sessions (up to 14 days) and workloads requiring specialized hardware or extensive resources. Users create capacity providers specifying EC2 instance types, such as GPU, memory-optimized, or compute-optimized, allowing flexible, scalable agent operation.
AgentCore handles provisioning, patching, scaling, and lifecycle management, simplifying operational overhead. This service runs in multiple AWS regions and incurs charges for both instance management plus EC2 usage. The runtime instances target teams needing orchestration of sustained or specialized AI agent workloads on a broad range of EC2 compute options.
Source & further reading
AgentCore runtime instances are now generally available ↗Source published 2026-08-06. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.