AIBrake
x402 API detects AI agent safety failures—loop detection, retry storms, and unverified deployments—in <50ms.
Trust signal
4 of 4 pillars
coverage 100%
Pillars
Signals
Evidence tiers, not an endorsement · trust model v0.1.0 (provisional) · CDP-lane settlement only (D3) · Trust API coming soon
On-chain traction
ERC-8004 agent registry
Not found in the ERC-8004 agent registry. This is the default state — absence is not a negative signal.
Identity & classification
Settlement volume
USDC settled on-chain · monthly
as of Jul 12, 2026
Endpoints(1)
Agent safety for paid AI agents — loop detection, model stop-loss, retry-storm catch, unverified-deploy block, premium-model-burn guard. AIBrake returns a decision (allow / warn / require_confirmation / block) for a proposed paid agent action — paid LLM call, deployment assertion, model escalation. Catches retry storms, unverified deploys, premium-model burn, objective drift, and budget breaches before the next expensive step. 98.0% LCR on published synthetic corpus, p50 0.004 ms.