Amazon Bedrock AgentCore is a set of AWS-managed infrastructure services for building, deploying, and operating AI agents in production, aimed at engineering teams that need enterprise-grade runtime, memory, identity, and observability primitives rather than a low-code agent builder.
Who it's for
AgentCore is built for developers and platform teams already working inside AWS who are building custom agents — with frameworks like LangGraph or CrewAI, or their own code — and need production infrastructure around them: secure execution, session memory, tool gateways, and identity management. It is not aimed at business users looking for a drag-and-drop agent builder; there's no no-code layer documented, and using it effectively assumes comfort with AWS services and API-driven development.
How it works
AgentCore is modular: a Runtime component executes agent code in isolated sessions, a Gateway turns existing APIs and Lambda functions into agent-callable tools, a Memory service persists short- and long-term context across sessions, an Identity service manages credentials and permissions, and Observability provides tracing and monitoring. It also includes managed browser and code-interpreter tools so agents can browse the web or execute code as part of a task. Because it's model-agnostic, teams can bring models from Amazon Bedrock or other providers rather than being locked into a single model family, and everything is accessible through AWS APIs and SDKs.
Pricing
AgentCore is billed on AWS's standard paid, consumption-based model rather than a flat subscription, and there is no published flat starting price or free tier in our data. Costs depend on usage across the individual services (Runtime session time, Gateway calls, Memory storage, and so on), similar to how other AWS services are metered. Teams should model expected usage and check AWS's current Bedrock AgentCore pricing page before committing, since consumption-based AWS pricing can be difficult to estimate without a usage forecast.
Strengths and trade-offs
The strength of AgentCore is that it solves the unglamorous but critical infrastructure problems of running agents in production — session isolation, memory, identity, and observability — as managed AWS services rather than things a team has to build themselves, and it plugs into whatever agent framework or model a team already prefers. Audit logging supports enterprise governance needs. The trade-off is that it's infrastructure, not an out-of-the-box agent product: there's no deployment outside AWS's own infrastructure, no no-code builder, and teams need engineering resources and AWS familiarity to get value from it. It competes less with consumer-facing agent tools and more with platform-level offerings like Microsoft's Azure AI Foundry Agent Service or Google's Vertex AI Agent Builder.