HUMANENDPOINT / GROUND TRUTH API
Real-world ground truth for AI agents.
HumanEndpoint gives autonomous agents a machine-readable way to obtain facts software cannot reliably observe itself. Request a verification, phone-sourced answer, photo observation or human judgement result; receive explicit terms; pay programmatically; and retrieve structured evidence.
OPEN API SPECVIEW CAPABILITIESWhat agents can obtain
Verify a lawful real-world fact and return structured observations with photographic evidence.
Get information that requires calling a business or organisation and return it as structured data.
Capture requested observations, images or measurements in the physical world.
Use bounded human interpretation for QA, research, AI-output verification and unusual edge cases.
Built for autonomous workflows
Agents can discover capabilities at /v1/services, check policy before committing, request a human-reviewed quote, accept terms, settle the service fee with x402 USDC on Base mainnet, poll job state and retrieve evidence protected by capability tokens.
Why ground truth matters
Software agents are excellent at operating on digital information. They are weaker when the answer exists behind a phone call, inside a physical location, in a human usability judgement, or in a real-world state that has not been published online. HumanEndpoint exposes those gaps as callable capabilities.
Start with HumanEndpoint
Agents should check /health, read /trust.json, discover capabilities at /v1/services, and use custom-human-task when an unusual lawful outcome does not match a named capability.