OpenA2A /specs
Open standards for AI agents · drawn out

Identity, trust, and authorization for AI agents.

One question repeats as agents call each other's tools and APIs: which agent is this, what can it do, and should I trust it? Five open specifications answer it, reusing the security the web already runs on: PKI, TLS, Certificate Transparency, and OAuth.

did:opena2a · AIP · ATX · ATP · AAP

Fig. 00 · the trust circuitrequest → current
agentDIDAIPATXATPAAPthe secret, sealed
named → proven → credentialed → logged → authorized
§01 · Why these specs exist

An agent is not a user, and not a server.

Humans log in with passwords and MFA. Servers prove themselves with TLS certificates. But an AI agent is a moving target: the same agent, with the same permissions, behaves differently depending on its prompt, its memory, and what it just read on a web page.

That breaks the assumptions behind OAuth, SAML and API keys. An agent can be talked into misusing a credential it legitimately holds. It can leak a secret simply by being asked nicely. And it can claim to be anyone.

OpenA2A treats agents as first-class cryptographic principals, with verifiable names, portable trust, and authorization that never puts a secret where a hijacked model could read it.

Identity
Who is this agent?
A cryptographic identity, not a self-declared name.
Identity
Is it really them?
Prove possession of the private key with a challenge.
Trust
Should I trust it?
A signed, portable, offline-verifiable credential.
Trust
Can I audit that?
Every trust decision in a tamper-evident public log.
Authorization
What may it touch?
Scoped access via a broker, the secret never reaches it.
Implementation
Where does it run?
AIM, the reference platform that issues & enforces it all.
§02 · See it decide

Break it, and watch which spec catches it

One request runs through all five gates. Every check starts valid. Flip any one to its failure state and the request is denied at the exact gate that owns that check, with the reason it exists.

Fig. 06 · run a request through the stack
agentcaller
DID
AIP
ATX
ATP
AAP
secretsealed
·Evaluating did:opena2a

Flip a condition

did:opena2aDoes the identity resolve?
AIPDoes it hold the key?
ATXIs the credential valid?
ATPHas it been revoked?
AAPIs the action in scope?

Every check defaults to valid. Set one to its failure state to see which spec catches it.

§03 · The core stack

Each layer stands on the one below it

Identity is the foundation. Trust is built on proven identity. Authorization is built on verifiable trust. AIM implements all three.

L3Authorization

How is trust turned into scoped access, safely?

L2Trust

How much should I trust it, and can I prove it offline?

L1Identity

Who is this agent, and is it really them?

Implementation
AIM

Issues and enforces every layer above. The reference platform.

Explore →
Each layer assumes the one beneath it · authorization needs verifiable trust · trust needs proven identity
§04 · The specifications

Read them in any order

Core protocol stack

did:opena2a0.1 (draft)
Decentralized Identifier Method

A W3C DID method where did:opena2a:<type>:<id> resolves over HTTP to a signed DID Document, so any verifier can fetch an agent's public key and trust endpoints.

like DNS + a certificate authority's directory
AIP1.0.0-draft
Agent Identity Protocol

An open standard for creating cryptographic agent identities, declaring capabilities, proving key possession via challenge-response, and computing a 9-factor trust score.

like The TLS handshake + an enterprise IAM directory
ATXv1.0 (spec) · v1.1 issued in production
Agent Trust eXtension

A signed, self-contained, 7-day credential, the TLS certificate for agents, encoding identity, scan results, capabilities and behavior, verifiable locally with no callback to any authority.

like A TLS / X.509 certificate
ATP1.0.0-rc1
Agent Trust Protocol

The protocol that issues, verifies, distributes and revokes trust assertions, recording every one in an RFC 6962 Merkle transparency log that anyone can audit.

like Certificate Transparency (CT) logs
AAP0.2.0-draft
Agent Authorization Protocol

An authorization layer where an agent emits an abstract grant:// reference and a local broker resolves scoped access, so no secret, token, or backend name ever enters the model's context.

like OAuth 2.0 Token Exchange, behind a broker
AIM1.0.0
Agent Identity Management

The reference platform that mints identities, issues credentials, runs 5-step authorization and keeps the audit trail, the working implementation of AIP, ATX, ATP, AAP and did:opena2a.

like The CA software + IAM server for agents

Threat, governance & measurement

The standards that define what can go wrong, what good behavior looks like, and how to measure and observe it all.

§05 · Nothing here is new, only new for agents

You already trust this exact design

Every time your browser opens a padlock, it runs this playbook. OpenA2A re-uses it, almost one-to-one, for agents.

The web you already trust
OpenA2A for agents
Certificate AuthorityDigiCert, Let’s Encrypt
Trust authorityATPissues credentials
TLS / X.509 certificateproves a server
ATX credentialATXproves an agent
Certificate Transparency logRFC 6962
Transparency logATPsame Merkle tree
OAuth token exchangescoped access
grant:// + brokerAAPsecret never seen
Same guarantees, same math, re-pointed from domains to agents.