Connect the landscape
Read metadata from an authorized, configured source. Keep platform scope, last-seen evidence and refresh failures visible.
AI AGENT REGISTRY & SYSTEM OF RECORD
Every AI agent.
One accountable record.
CatProof brings configured agent inventories into one federated system of record. Link source identities, name the business owner, and keep your decisions as source metadata changes.
Early access · No credit card · Private registry per account
Curious by nature. Accountable by design.
WHAT IS CATPROOF?
A home for your agent inventory.
A clear answer to “who owns it?”
CatProof is an early-access AI agent registry for teams working across multiple platforms. It separates business ownership and lifecycle decisions from vendor metadata, while retaining the source identity and freshness behind each record.
FROM SCATTERED TO ACCOUNTED FOR
Connect. Assign. Keep the evidence.
Name the owner
Link source identities to a canonical agent record. Assign the business owner and local lifecycle deliberately.
Preserve the decision
Refresh observations without overwriting ownership, review notes or lifecycle. Inspect source changes and decisions in history.
THE RECORD THAT TRAVELS WITH THE AGENT
Two source identities.
One business owner.
One logical agent may appear in multiple registries. CatProof preserves each source ID while keeping the business record separate.
Try the ownership workflow →ILLUSTRATIVE RECORDFictional
Invoice triage
Accountable owner Jordan Lee · Finance
Canonical lifecycle Active
Snowflake CortexA2A Agent Card
Source identities linked deliberately. This is demo data.BUILT FOR A MIXED AGENT ESTATE
Your platforms. A common inventory.
Snowflake Cortex AgentsMicrosoft Copilot StudioAWS Bedrock AgentsGoogle Agent RegistryA2A Agent Cards
Adapters read configured scope, not every agent in an organization. Authentication, permissions and validation status vary by connector.
See connector coverage and limitations →What is a federated AI agent registry?
A federated AI agent registry reads inventories from configured sources and relates their source identities to common agent records. Source systems remain responsible for their own definitions and runtime state.
Does CatProof control or run agents?
No. CatProof reads source metadata and stores local ownership, lifecycle and review decisions. Setting a canonical lifecycle to suspended does not stop the agent in its vendor platform.
Does discovery find every agent automatically?
No. Discovery covers the sources, regions, projects and permissions configured for each connection. Agents outside that scope may be absent. Failed refreshes retain prior evidence with a stale indicator.
Can I try it without provider credentials?
Yes. The public interactive demo uses fictional data. A new account can also register manual records and connect public A2A or MCP endpoints. Protected providers require authorized credentials provisioned for the registry.
What is included in early access?
Personal signup, a separate registry per account, manual agent records, connector configuration, ownership and lifecycle fields, explicit identity links, refresh history, team roles, revocable invitations and session controls. Email verification, password recovery, invitations, SSO and billing are not connected yet.
Are all connectors verified in customer environments?
No. The catalog distinguishes implementation and fixture or contract testing from live-account validation. Real coverage must be checked against the authorized tenant and configured scope.
START SMALL. MAKE OWNERSHIP CLEAR.
Give your first agent a proper record.
Create an empty private registry, then register an agent or connect an authorized source.