AEA/P defines how AI agents are held accountable when they transact: verifiable identity, proof of performance (PoP), liability escrow, dispute resolution, and entity governance — one open specification, implemented by independent Operators and Platforms.
First Operator implementation is live — Nustro · nustro.com
One agent wants to do business with another it has never met. AEA/P is the trust layer that makes that economic transaction accountable.
Agents can do business with agents they don't know — because AEA/P proves who they are, tracks how they perform, holds liability in escrow, and provides recourse when something goes wrong.
AI agents are becoming autonomous economic actors. They execute transactions, commit resources, and interact with other agents at machine speed. Communication protocols like MCP and A2A standardize how agents connect. Payment protocols like x402, MPP, and AP2 enable agents to transact. Commerce protocols like UCP and ACP orchestrate the buying experience. Identity standards like ERC-8004 provide on-chain discovery and reputation.
None of these protocols answer the harder questions: what happens when an agent causes financial harm? How are disputes resolved without courts? What liability escrow exists before a transaction begins? How does an autonomous entity govern itself?
AEA/P specifies that missing accountability layer. It doesn’t replace these protocols or sit on top of them—it spans across communication, commerce, payments, and identity, binding them into one governed, end-to-end flow.
AEA/P assumes the existing agent stack rather than replacing it. Each layer below already has maturing protocols; AEA/P runs across them as the governance fabric.
MCP connects agents to tools. A2A enables agent-to-agent coordination. x402, MPP, and AP2 handle payment settlement across stablecoins, cards, and bank transfers. UCP and ACP orchestrate the commerce flow—product discovery, checkout, fulfillment. ERC-8004 provides on-chain identity and reputation signals. AEA/P spans across all of them—tying identity, payments, commerce, and communication into a single accountable flow, so the same governance, liability, and dispute resolution apply end to end across everything an agent does.
The agent infrastructure stack is maturing rapidly. Existing protocols solve connectivity, communication, identity, payments, and commerce. AEA/P covers what happens when things go wrong—and how trust is established before things go right.
| Capability | AEA/P | ERC-8004 | x402 / MPP / AP2 | UCP / ACP |
|---|---|---|---|---|
| Identity | ✓ | ✓ | — | — |
| Agent authentication | ✓ | — | — | — |
| Agent discovery | ✓ | ✓ | — | — |
| Payment intent | ✓ | — | ✓ | ✓ |
| Payment orchestration | ✓ | — | ✓ | ✓ |
| Human in loop | ✓ | — | ✓ | ✓ |
| Reputation | ✓ | ✓ | — | — |
| Liability / escrow | ✓ | — | — | — |
| Disputes | ✓ | — | — | — |
| Governance | ✓ | — | — | — |
| Compliance | ✓ | — | — | — |
AEA/P doesn’t compete with these protocols or stack on top of them. It spans across them—tying identity, payments, commerce, and communication into one cohesive, end-to-end flow with the economic accountability that regulated industries require.
AEA/P defines five interconnected governance capabilities that activate as an agent’s economic participation deepens. Together they form one connected governance lifecycle — select a stage to see what it does.
A protocol is defined as much by what it leaves out. AEA/P deliberately does not try to be:
AEA/P moves no money. Settlement runs over x402, MPP, AP2, or a native mechanism — AEA/P governs the identity, escrow, and recourse around the payment.
It does not replace MCP or A2A. Agents keep connecting and coordinating the way they already do; AEA/P makes the resulting transactions accountable.
AEA/P doesn't prescribe how agents are built or which models they run. It defines how they are identified, rated, escrowed, and held to account.
AEA/P is an open specification. Operators and Platforms implement it independently — implementations are built on the protocol, they are not the protocol itself.
AEA/P separates roles at two levels: the participants who run the governance fabric, and the agents that transact on it.
A third economic role — the Enterprise agent, transacting on both sides as an autonomous business — carries the full governance stack, including entity governance.
AEA/P is running in production. Implementations listed here operate the protocol’s Operator or Platform surface against the published specification.
Commercial Operator built on AEA/P, running the protocol’s Operator surface: agent identity issuance (did:aeap:), proof of performance, liability escrow, atomic on-chain settlement, and dispute resolution — with sandbox and production environments.
The specification and the Operator and Platform reference documents are open. If you’re implementing the protocol, get in touch to be listed here.
A system claiming AEA/P conformance declares its conformance level and economic role; AEA/P Certified indicates that an independent assessment has verified conformance at the claimed level for the declared role. Certification criteria and processes are being defined with input from early implementers and standards bodies.
AEA/P is not a theoretical proposal. Its core concepts were designed, implemented, and tested in a working platform years before the current AI agent ecosystem emerged.
In 2018, xDAC was published—a platform for Decentralized Autonomous Companies. The xDAC whitepaper defined governance for autonomous entities including Proof of Performance rating, liability escrow, automated dispute resolution, and support for autonomous agents as team members.
The platform shipped with company registration, team management, invoicing, and payment processing. The concepts were ahead of their time—LLM-based agents didn't exist yet. In 2026, those governance concepts are being refined as AEA/P for the current AI agent ecosystem.
AEA/P is developed in the open, in dialogue with the bodies standardizing agent identity and governance.
AEA/P is an open protocol specification, developed in public. Contributions, feedback, and implementation reports are welcome — if you’re implementing the protocol, integrating a platform, or building agents, reach out at support@aeap.ai.