Who vouches for this AI agent?
A neutral verification layer: any organization can independently verify another company’s AI agents. Who they are, what they may do, and whether they’ve been revoked.
Trust breaks when one company’s agent acts on another’s systems
Identity platforms answer who can access your systems. Nobody answers why you should trust another company’s agent.
From agent identity to cross-organization verification
Five steps. Agent identity and revocation can be independently verified by any organization, with the record kept on a public chain. As SSL did for websites, every agent that crosses a company line will need a check anyone can verify.
- ProvisionUnique DID anchored on Ethereum
- CredentialAgent self-signs a verifiable credential
- Policy checkReal-time enforcement at the OIDC layer
- AWS accessShort-lived STS credentials issued
- VerifyAny organization checks the agent, no account
What another organization can check
Three things every counterparty needs to know. Attribution and KYB linkage are next.
Who the agent is
Cryptographic proof that the caller controls the agent’s identity key (its DID), and that it is registered, with who registered it and when. Live on testnet.
What it’s authorized to do
A credential stating the agent’s permissions. The permissions part is live. Spending limits and scope per agent are next, not built; enforced outside the agent, they cap the damage if an agent is manipulated.
Whether it has been revoked
Public, final revocation, with block and time shown. Live on testnet.
What we’re building next
Attribution: access events anchored as Merkle roots, so a disputed action can be traced to an agent. KYB linkage ties an agent to a verified business.
Status: identity and revocation run on testnet; permissions are live; limits and scope, attribution and KYB linkage are next. Meashra caps damage and makes agents traceable; it does not detect fraud or stop an agent being manipulated.
Check an agent yourself
The public verifier reads Meashra’s public registry (Ethereum Sepolia testnet) directly. It calls no Meashra server and needs no account or key of ours. It shows whether an agent is registered, whether it was revoked, and whether a presented DID document matches the chain.
Open the verifierDesign partners
We are working with a small number of design partners to shape how agents get verified. We would like to hear from regulated organizations (financial services, fintech and crypto platforms, payment providers) that receive agent traffic from other companies, and from companies running AI agents that need their partners to trust them.
What you get: direct access to the founders, a say in the credential format, authority limits and verifier, and early access to the live beta (Ethereum Sepolia testnet).
What we ask: 5 minutes to tell us how you check another company’s agent today, and what you cannot check.
How it is meant to work: the organization running an agent enrolls it and pays per agent; the organizations that deal with it verify for free. We are looking for the first counterparties to ask their agent partners to enroll.