Teleport Debuts Delegated Identity and LLM Proxy in Beams Public Beta

Teleport, an identity company, said its Beams trusted, ephemeral agent runtime added two capabilities in public beta that target identity and access controls for AI agent workflows. The change addresses how agent behavior is governed when agents interact with inference endpoints and infrastructure resources.

Teleport framed the update around an agentic workflow problem: identity, access control, and audit are not applied at the points where an agent is instructed and where it can reach the resources it is allowed to use. It said the capabilities work within the identity and zero trust plane that governs privileges and infrastructure resources, including production databases, cloud APIs, and internal services.

For LLM Proxy, Teleport described an enforcement layer positioned between an agent and its inference endpoint. It said every request and response is inspected and written to an audit log, and that controls are enforced through a per-Beam allow list of resources that determine which agents can reach which inference endpoints and under what conditions. It said the audit log serves as a tamper-resistant record of what each agent was told to do.

Delegated Identity, Teleport said, lets a human operator or an agent define and assign the permissions an agent will carry, with scoped privileges for a specific task delegated by human or another agent. Teleport said agents receive delegated identity rather than broad credentials or standing access, and that all activity is recorded, analyzed by peer agents, and tied to identity and task. It also said the runtime makes least-privilege access a property of each agent, with blast bounds if an agent is compromised or behaves unexpectedly.

Forward-looking: Teleport said LLM Proxy and Delegated Identity are available now in the public beta of Beams.

Provided by Globe Newswire on behalf of Teleport. Click to read original content.