Most agent frameworks today make it easy to spin up a new AI agent, but they make it hard to answer a basic question: who is this agent, and what is it allowed to do? Industry conversations around “Know Your Agent” (KYA) are starting to frame this as the agent-side counterpart to traditional KYC.
With AOM Agent Kits, we add a simple registration step so that every agent can have a stable identity, an accountable owner, and a clear purpose from day one.
What “Know Your Agent” Means
For humans, KYC is about verifying people before you trust them with money or data. For AI agents, KYA is about three things:
- Who owns this agent (person or organization).
- What surfaces and data it is allowed to touch.
- How to revoke or rotate its access when something changes.
If you can’t answer those questions, you effectively have anonymous scripts roaming your product. KYA is the process of turning those scripts into named, accountable agents.
From anonymous scripts to registered agents
By default, most agents look like anonymous scripts running somewhere in your infrastructure. The AOM Agent Kits add one missing piece: a lightweight registration step that gives each agent a key, an owner, and a declared purpose.
On the Register Agent page, we ask for:
- Owner name and email – so there is a real, accountable owner behind the agent.
- Agent name – a stable identifier you can use in logs, dashboards, and alerts.
- Purpose and organization – a short description of what the agent is supposed to do and who it acts on behalf of.
Registration is free, and you can still download and run AOM Agent Kits without registering. Registration is what upgrades a script into a trackable, accountable agent.
How the agent key works
When you complete the form, AOM Tools issues an agent key that you add to your aom.kya.agent.json file in the AOM Agent Kit.
On startup, the kit validates this key via the backend validator and only proceeds if the key exists and is not expired. If the key is missing, invalid, or expired, the agent exits early and returns a clear error in the AOM output so your runtime can prompt for re‑registration or rotation.
This has a few important KYA properties:
- Every run is tied to a registered agent identity.
- If your key expires or is revoked, the agent exits on startup instead of silently running with stale credentials.
- You can build logging and analytics around agent IDs (and owners) instead of scattered script names and hostnames.
The registration flow now also shows how many agents are registered, which gives a public sense of how many agents are in the ecosystem.
Why this matters for surfaces and policies
The Agent Object Model (AOM) already gives agents a structured view of the surfaces they interact with, using JSON definitions and automation policies (`forbidden`, `ready/allowed`, and `open`). When you combine that with registered agent identities, you can start enforcing rules like:
- “This agent can only operate on surfaces that are `automation_policy: open`.”
- “This class of agents is restricted to surfaces belonging to this organization.”
This is where KYA gets powerful:
- Surfaces declare where automation is allowed.
- Agents declare who they are and what they’re for.
- Your governance layer can match the two, instead of letting arbitrary scripts roam across your product.
Getting started with KYA in AOM
You can start experimenting with KYA in AOM today:
- Download the AOM Agent Kits from the Downloads → Agent owners section.
- Register your first agent on the Register Agent page to get a key and identity.
- Wire the key into your agent startup and log the agent ID for every action your agent takes.
You don’t have to redesign your whole stack to get KYA benefits; you just need a consistent way to name agents, assign owners, and validate keys before they touch real surfaces.
Over time, we expect KYA to grow into richer features: more detailed purposes, org‑level controls, and deeper analytics for understanding how your agents behave in the wild. The registration and validation flow in AOM Tools is the starting point.