Trust that grows with every run.

A console for teams who let AI agents act on their behalf: what ran, what is waiting for approval, and how to step in.

The question

Relay lets operations teams hand repetitive work to AI agents. The hard part isn’t what the agents can do. It’s whether people feel safe letting them do it.

The idea

Make every action legible. Each run reads like a short story: what the agent saw, what it decided and why, and where a person can step in.

The system

A calm interface built around one pattern, the approval card, which carries the agent’s reasoning, the policy it followed and two clear choices.

The motion

Motion stays functional. Steps tick through as they complete, and waiting items pulse once, gently, instead of demanding attention.