Hey PH — I’m one of the builders behind AgentOracle, and I want to be honest about what we built and why. I kept working with AI agents that search for information and act on it — with no signal at all about whether to trust the result. An agent getting back a 0.95-confidence answer and a 0.31-confidence answer should behave completely differently. But every API I looked at just returned text and left that judgment to the developer. So we built the confidence score first. Everything else — the /compare endpoint, the min_confidence parameter, the structured JSON — came from one question: what does a research result need to return for an agent to make a good decision? We went x402 because it fits agents perfectly. Pay per query, on-chain, no account overhead. Live on Base, SKALE (gasless), and Stellar (first Stellar x402 payment was April 6). Today Exa announced x402 web search — exciting for the ecosystem. AgentOracle sits on top of that retrieval layer — the trust and scoring layer that makes retrieved information safe to act on. Free preview live now: agentoracle.co/preview — 20 req/hour, no auth. Try it and tell me what breaks. I’m here all day.

AgentOracle

Research API for AI agents — confidence scored, x402 native

Research API for AI agents — confidence scored, x402 native