AWS cuts agent setup to three calls

- Amazon Web Services added a managed harness to Amazon Bedrock AgentCore on April 22 that lets developers define and run agents in three API calls. - AWS said each AgentCore session runs in a dedicated Firecracker microVM, with isolated CPU, memory, filesystem, and sanitized memory after shutdown. - The launch adds a command-line interface and coding skills as AWS chases enterprise agent deployments over model lock-in. (aws.amazon.com)

Amazon Web Services added a managed harness to Amazon Bedrock AgentCore on April 22 that lets developers stand up a working agent in three API calls. (aws.amazon.com 1) (aws.amazon.com 2) The harness replaces orchestration code with configuration: developers specify a model, system prompt, and tools, then invoke the agent through AgentCore. AWS said the feature is in preview. (aws.amazon.com 1) (aws.amazon.com 2) AgentCore is the runtime layer underneath that setup. AWS documentation says each user session runs in its own Firecracker microVM with isolated CPU, memory, and filesystem resources, and the microVM is terminated with memory sanitized when the session ends. (docs.aws.amazon.com 1) (docs.aws.amazon.com 2) AWS says that isolation is meant to stop one agent session from leaking state, credentials, or permissions into another session when tools act on a user’s behalf. The company also says the harness assumes an AWS Identity and Access Management role owned by the customer, using least-privilege access. (docs.aws.amazon.com) (docs.aws.amazon.com) The pitch is not only speed. AWS is packaging agent deployment as governed infrastructure, with security boundaries, session management, scaling, and observability handled by the platform instead of custom glue code. (docs.aws.amazon.com) (aws.amazon.com) AWS is also pushing model flexibility in the same release. Its documentation says AgentCore Runtime works with models from Amazon Bedrock, Anthropic Claude, Google Gemini, and OpenAI, and the service is described as working with any framework and model. (docs.aws.amazon.com) (docs.aws.amazon.com) The April 22 update included two other pieces: an AgentCore command-line interface for prototyping and deployment, and AgentCore skills for coding assistants. AWS said the command-line interface, harness, and skills carry no additional charge beyond the underlying resources used. (aws.amazon.com) (aws.amazon.com) AWS said the managed harness is available in preview in four regions: US East (N. Virginia), US West (Oregon), Europe (Frankfurt), and Asia Pacific (Sydney). The command-line interface and persistent agent filesystem are available globally. (aws.amazon.com) (aws.amazon.com) The release leaves AWS selling more than a model endpoint. It is selling the scaffolding around agents — the session boundary, tool access, and deployment path — in a market where companies are trying to move prototypes into production. (aws.amazon.com) (siliconangle.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.