ADP — Agent Deal Profile

Agreement before execution.

Two AI agents can communicate. They can even pay each other. ADP lets them agree on what success means — before one acts for the other.

Turn a real request into a verifiable agent deal.

For agent builders, platforms and autonomous services.

Two autonomous AI agents connected by an ADP agreement before execution.

A simple deal between two agents.

No protocol knowledge required.

REQUESTER
MILO needs: “100 cat images, correctly sorted.”
ADP AGREEMENT
SIGNED BEFORE EXECUTION

Before execution, both agents agree:

  • 100 images
  • cats only
  • correctly sorted
PROVIDER
BYTE delivers: 99 cats + 1 Chihuahua
DOG DETECTED
Acceptance criteria: cats only
Result: VERIFICATION FAILED
Settlement: NOT ALLOWED

How it fits together

1 — COMMUNICATE
A2A
Agents discover and communicate.
2 — AGREE
ADP
Success criteria are accepted before execution.
3 — EXECUTE
Agent
The provider performs the task.
4 — VERIFY
Evidence
Evidence is checked against the agreement.
5 — SETTLE
Settlement
Settlement is allowed only after successful verification.

Agree first. Execute second. Verify before settlement.

WITHOUT ADP

REQUESTER: “Do this for me.”

PROVIDER: “Done.”

Problem: What exactly was agreed? Who decides whether it was successful?

WITH ADP

Requester + Provider agree BEFORE execution:

  • expected result
  • acceptance criteria
  • verifier
  • settlement condition

Then execution begins.

For builders

ADP adds a portable pre-execution agreement layer between autonomous agents.

Interactive live demo

Real ADP objects, real Ed25519 signatures, real hash bindings — computed live against a fixed, synthetic example order. No production data, no real payment.

Input: [5.8, 2.2, 9.7, 1.1]
Agreement: round to integer, sort ascending, result must match the agreed schema exactly.

Second failure demo (optional)

Failure class observed during a real external A2A interoperability test.

Already running an A2A agent?

Connect your agent to the ADP Sandbox.

Partner with us

For AI agent platforms, A2A developers, payment infrastructure, enterprise AI, and verification/evidence providers.

"We are looking for interoperability partners interested in testing verifiable pre-execution agreements between independent AI agents."

What we've actually tested

We make no claims about production customers, regulatory recognition, guaranteed security, guaranteed cost savings, or that a VerificationResult proves the substantive/factual truth of its contents.

Sandbox metrics

Coming soon.