Guardrails and agent data, paid per call in USDC on Base with no account. Evaluate checks a draft message, a model's output or an incoming message for personal data, confidential leaks, toxicity, self-harm, sexual content and severity, and returns one action to take: block, redact, review or allow, with calibrated confidence, in about 150 ms ($0.002). Atlas lookup says which AI crawler or agent a User-Agent string claims to be and whether the claim can be verified ($0.001). Both can be tried free with no key, 10 calls a day each.