Generative AI platform
Amazon Bedrock AgentCore
Agentic platform for building, deploying, and operating AI agents securely at scale with any framework and foundation model.
Key points
- Runtime hosts agents serverlessly with session isolation.
- Memory provides short-term and long-term memory for context-aware agents.
- Gateway turns APIs and Lambda functions into MCP-compatible tools for agents.
- Identity manages agent authentication with existing identity providers, and Policy enforces rules on every tool call.
- Observability traces and monitors agent steps, and Evaluations measures agent quality.
- Works with open-source frameworks such as Strands Agents, LangGraph, and CrewAI, and with models in or outside Amazon Bedrock.
When to use it
- Choose AgentCore to build agents that plan multi-step tasks and call tools or APIs.
- Use Gateway to expose existing Lambda functions or APIs to agents as tools.
- Use Identity and Policy to control what an agent may access and do.
Exam tips
- AgentCore is the recommended path for new agents; Bedrock Agents (now Agents Classic) closed to new customers on July 30, 2026.
- Agents take actions; Knowledge Bases retrieve context; Guardrails filter content.
- AgentCore Identity and Policy in AgentCore appear in the exam guide objective on securing AI systems.