The Bedrock AgentCore SDK is AWS’s official Python SDK for turning a locally developed AI agent into a production application running on Amazon Bedrock AgentCore. Published on GitHub with about 780 stars under the Apache-2.0 license and updated in October 2026, it provides framework-agnostic building blocks for deployment.
Amazon Bedrock AgentCore is AWS’s managed infrastructure for deploying and operating agents securely at scale, with any framework and any model. The SDK wraps a local agent — for example one built with the Strands Agents framework — in an application entrypoint, then connects it to AgentCore services: a managed runtime for hosting, memory for retaining context, identity and authentication for controlled access, and tool integrations. The services are composable, so developers can adopt only the pieces they need while AWS handles the underlying scaling and reliability.
The SDK’s design follows the shape of the AgentCore platform itself. Developers annotate a handler as the entrypoint, and the framework takes care of exposing the agent to the managed runtime; from there, composable services add production capabilities — session memory so agents remember context, identity controls so access is authenticated and scoped, and tool gateways so external APIs can be called safely. Sample applications in the repository show end-to-end deployments, and AWS maintains related Boto3 and runtime SDK documentation alongside it.
The SDK itself is free and open source under the Apache-2.0 license. Running agents on Amazon Bedrock AgentCore is billed under AWS’s pricing for that managed service.
The SDK is for Python developers and teams who have an agent working locally and need enterprise-grade hosting, security controls, and scaling on AWS instead of operating agent infrastructure themselves.