Amazon Web Services (AWS) announced new enhancements for Amazon Bedrock AgentCore at AWS re:Invent in Las Vegas, focusing on improved security, continuous evaluation, and advanced learning capabilities for AI agents.
The new Policy in Amazon Bedrock AgentCore introduces real-time, deterministic controls that operate independently of the agent’s code. This feature is designed to actively block unauthorized agent actions, enhancing security and governance for AI agent deployments.
AgentCore Evaluations provides developers with tools to continuously assess the quality of an agent based on its behavior. This allows for ongoing inspection and refinement, supporting the development of more reliable and effective AI applications.
Additionally, AgentCore Memory introduces episodic functionality, enabling agents to learn from past experiences. This capability is intended to improve decision-making processes over time by allowing agents to retain and apply insights gained from previous interactions.
Organizations spanning various sectors, including Amazon Devices Operations & Supply Chain, Archera, Cohere Health, Cox Automotive, Druva, Heroku, Natera, NTT DATA, MongoDB, PGA TOUR, Pulumi, Thomson Reuters, Workday, Snorkel, Swisscom, and S&P Global Market Intelligence, utilize AgentCore to accelerate their AI agents into production environments.
