Verify AI agent reasoning quality

Adversarial multi-model verification: independent evaluators examine the reasoning, a separate red-team model challenges their verdicts, and a synthesizer produces the final verdict. Returns ALLOW, BLOCK, or UNCERTAIN with confidence score and JWKS-signed attestation.

post/v1/check

Request

  • Base URL: https://api.thoughtproof.ai
  • URL: https://api.thoughtproof.ai/v1/check
  • Auth: API key in header X-API-Key

Request body

claimstring required

The decision or reasoning to verify

stakeLevel'low' | 'medium' | 'high' | 'critical'
domain'financial' | 'medical' | 'legal' | 'code' | 'general'
speed'fast' | 'standard' | 'deep'

Verification depth. Pricing: fast=$0.008, standard=$0.02, deep=$0.08 (Base USDC via x402)

agentIdstring

Optional caller/agent identifier for per-agent verification history (also accepted via X-Agent-Id header; body wins). Powers GET /v1/stats/{agentId} and /v1/trustscore/{agentId}.

Response

Verification result

verdict'ALLOW' | 'BLOCK' | 'UNCERTAIN'
confidencenumber
objectionsstring[]
durationMsinteger

Example response

{
  "confidence": 0.82
}

Changes

No recorded changes to this operation across all 1 revision of this API.