TRY ADP
Agents can communicate. Agents can execute. ADP defines what both sides agreed to and checks the delivery against that agreement.
ADP does not do the work. It defines what must be delivered, records the agreement, and verifies the delivery. Not a scripted demo, not a random PASS. Not proof the verifier is always honest, not proof every semantic claim is true, not autonomous settlement.
Run a real tamper action against the real stored delivery/spec and re-verify with the exact same verifier logic.
DEFINE WHAT COUNTS AS SUCCESS
PROVIDER AGENT -- CONCEPTUAL FLOW
Full sandbox API documentation and a runnable example external agent (a genuinely separate OS process, talking only to the public API below) are described on the developer sandbox docs page.
POST /api/playground/external/deals
GET /api/playground/external/deals/{deal_id}
POST /api/playground/external/deals/{deal_id}/ack
POST /api/playground/external/deals/{deal_id}/delivery
GET /api/playground/external/deals/{deal_id}/verification