Agent Registry
A discoverable catalog of agents — internal and external — with their Agent Cards, capabilities, trust scope, and operational health, so other agents and humans can find and call them.
Production AI is not a prompt. It is a system of context, tools, permissions, traces, evals, and feedback loops.
What it is
An Agent Registry is the equivalent of a service registry but for AI agents. It lists which agents exist in your organization (or across organizations you trust), what each one does, who owns it, what trust scope it has, and whether it's currently healthy. Client agents and integration code consult the registry to discover counterparts at runtime.
Why it matters
Once you have more than a handful of agents, hardcoding who-talks-to-whom becomes unmanageable. A registry is how you onboard, deprecate, scope, and version agents the same way you would APIs — with the additional dimensions of trust and capability that agents bring.
How it works
Each agent registers an Agent Card (capabilities, schemas, trust requirements). The registry handles authentication, capability search, health checks, deprecation paths, and tenant scoping. In an enterprise context, the registry often sits alongside the MCP Tool Registry — tools for what an agent can use, agents for what an agent can collaborate with.
Related resources
A structured description of what an AI agent can do, how to talk to it, and what trust it requires — the public face of an A2A-compatible agent.
A protocol for agents to discover, describe, and call each other directly — Google's open spec for letting independent agent systems collaborate without each one becoming a tool for the other.
A governed catalog of every tool an AI agent can call — your APIs, your databases, your internal systems — with typed schemas, permission scopes, audit trails, and the standard protocol (MCP) that turns 'we exposed it to the LLM' into 'we know exactly who called what when'.
A pattern where a supervisor agent delegates parts of a task to specialized sub-agents — each with narrower scope, focused tools, and its own success criteria.