Amazon Web Services announced the release of Loom, an open‑source platform designed to help enterprises build and deploy AI agents with robust security controls and governance frameworks on the AWS cloud. Loom integrates with Amazon Bedrock AgentCore and AWS Strands Agents, delivering lifecycle management for AI agents at scale.
The platform enforces three mandatory tags on all deployed resources and allows custom tagging for cost attribution and governance. It offers automated resource tagging, role‑based and attribute‑based access controls, and pre‑validated configuration blueprints that streamline agent deployments. Access control operates through a two‑dimensional system that combines role types with group tags to restrict user permissions.
Deployments are configuration‑driven rather than relying on runtime code generation, with behavioral guidelines and security credentials managed via AWS Secrets Manager. Loom supports low‑code deployments using pre‑written Python agents and no‑code deployments through AgentCore’s managed harness. OAuth2 configurations and token‑exchange protocols propagate user identity through the agent request chain.
Loom integrates with the AWS Agent Registry, currently in public preview, to manage agent and tool records and enforce governance review before production deployment. Human‑in‑the‑loop approval workflows are implemented using the Strands Agents hook framework and Model Context Protocol elicitations for sensitive actions requiring manual review.
The project is available through AWS Labs on GitHub, inviting community contributions, and is positioned for platform engineering teams building applications with fully managed AWS services. The article was generated with AI assistance and reviewed by an editor.