Vangrid makes physical data capture a paid task for software agents
Its documented bounty flow connects USDC funding to a human capture and a review step. The procurement boundary matters as much as the payment rail.
Vangrid’s documentation, last modified September 30 and checked on October 25, describes agents funding location-capture bounties in USDC on Base or Arc through x402. People film the requested space, and the requester reviews a submission before receiving the reconstruction. Its public board lists accepted and cancelled jobs; this is a documented workflow, not a newly verified launch.
The interesting separation is between paying to commission work and accepting the delivered result. Our analysis is that agent procurement needs both boundaries: a funded request can still fail to produce usable data. A payment receipt cannot establish coverage, reconstruction quality or permission to record a location.
That distinction is especially important for physical AI. The supplier’s documentation describes content fingerprints anchored on Base, but provenance and fitness for a training task answer different questions. A buyer would still need acceptance criteria for the scene, capture conditions and rights to reuse the material.
For regional builders considering similar services, the reusable idea is a clear handoff from software to a human task and back to review. The design separates commissioning work from accepting a physical deliverable; payment integration alone cannot establish data quality.