Many humans, many agents
Agent registry
An agent registry is an inventory of every agent in an organization recording its name, purpose, owner and escalation contact, the surfaces it runs on, what it may read and write, the identity it runs under and the data it touches, reviewed on a fixed cadence.
Control sits with the person
When to use it
Start a registry when your organization deploys its second agent. It costs much less to start one early than to build one later for agents that are already running.
What it prevents
Agent sprawl, named by Gartner in April 2026: old agents accumulate, dependencies go unmapped, and nobody can say which version is running.
Where it stands
Registries are a standard part of enterprise agent governance. They usually sit in a control plane outside the build and orchestration layers.
This entry is our own design position. It does not cite a released product or a published paper.
Related patterns: Agent identity · Permission parity
Worked through as a deck: Your newest teammate has a spend cap