VerityLayer check · safety
← all VerityLayer checks
Action guardrail for AI agents
An independent allow / review / block before your agent does something irreversible.
What it does
Before your agent spends, sends, publishes, or runs a destructive command, call the guardrail and only proceed on allow. Fail-closed: it will not allow what it cannot justify as safe.
decisionriskreasonsconcernssafer_alternative
| Tier | Price (USDC/Base) | Depth |
|---|---|---|
| quick | $0.02 | fast |
| standard | $0.08 | fast |
| pro | $0.20 | web-grounded |
curl -X POST https://suite.veritylayer.dev/check/quick \
-H "content-type: application/json" \
-d '{"action": "…", "context": "…"}'
# unpaid → HTTP 402 (x402 challenge); pay $0.02 USDC on Base and retry
Every verdict is signed. This check returns an Ed25519-signed receipt — portable proof of exactly what VerityLayer said. See it, and break one yourself →