
Amazon Bedrock AgentCore Policy is now generally available, enabling organizations to enforce centralized, fine-grained security controls over agent-tool interactions through natural language policies without modifying agent code.


<div><p>This article announces the general availability of Policy in Amazon Bedrock AgentCore, a new feature providing centralized controls for agent-tool interactions and governance.</p><ul><li>Centralized, fine-grained controls for agent-tool interactions outside agent code</li><li>Security teams define tool access and input validation rules without code changes</li><li>Natural language policy authoring automatically converts to Cedar policy language</li><li>AgentCore Gateway intercepts and evaluates agent-tool traffic against policies</li><li>Ensures agents operate within defined organizational parameters</li><li>Available in thirteen AWS regions across US, Asia Pacific, and Europe</li></ul><p>Policy in AgentCore enables organizations to enforce governance and security controls over AI agents without modifying application code, improving compliance and operational oversight.</p></div>


Related articles