{"uid":"cap_x90rZzt8oL2Z5qIF_ciXz","slug":"thoughtproof-sentinel-api-cefae7b5","name":"ThoughtProof Sentinel API","description":"Pre-execution verification checkpoint for autonomous AI agents. Multi-model cascade evaluates agent reasoning before irreversible actions. Returns ALLOW, BLOCK, or UNCERTAIN with structured per-step objections and optional EAS on-chain attestation.","url":"https://sentinel.thoughtproof.ai/sentinel/verify?utm_source=zero.xyz","method":"POST","headers":{},"bodySchema":{"type":"object","required":["mode","claim","evidence"],"properties":{"id":{"type":"string","description":"Optional verification ID (auto-generated if omitted)"},"mode":{"enum":["handoff","plan_revision","memory_write","output_synthesis","trade_execution","trade_reasoning","action_authorization"],"type":"string","description":"Verification mode. handoff: agent-to-agent delegation. plan_revision: plan change review. memory_write: persistent memory gate. output_synthesis: final output check. trade_execution/trade_reasoning: financial decision verification. action_authorization: deterministic gate + LLM."},"tier":{"enum":["checkpoint","standard"],"type":"string","description":"Verification tier. checkpoint: nano solo ($0.005). standard: 2-model cascade ($0.008, default)."},"claim":{"type":"string","description":"The agent decision or action to verify"},"mandate":{"type":"object","description":"Optional machine-readable authorization mandate for action_authorization mode. Enables deterministic gate checks (amount limits, recipient allowlist) before the LLM."},"evidence":{"type":"string","description":"Context, reasoning trace, or market data supporting the claim"},"gateMode":{"enum":["shadow","enforce"],"type":"string","description":"Deterministic gate rollout stage (action_authorization only). shadow: logs violations but does not change verdict. enforce: violation forces BLOCK."}}},"responseSchema":{"type":"object","required":["id","verdict","confidence","reasoning","objections","mode","tier","meta"],"properties":{"id":{"type":"string","example":"req_m1abc_x9f2kq"},"gate":{"type":"object","properties":{"mode":{"enum":["shadow","enforce"],"type":"string"},"enforced":{"type":"boolean"},"violations":{"type":"array","items":{"type":"object","properties":{"type":{"type":"string"},"message":{"type":"string"}}}},"wouldBlock":{"type":"boolean"}},"description":"Deterministic authorization-gate result (action_authorization mode only)"},"meta":{"type":"object","properties":{"duration_ms":{"type":"integer","example":1200},"models_used":{"type":"array","items":{"type":"string"},"example":["serv-nano","serv-swift"]},"verified_at":{"type":"string","format":"date-time"}}},"mode":{"type":"string"},"tier":{"type":"string"},"verdict":{"enum":["ALLOW","BLOCK","UNCERTAIN"],"type":"string","description":"ALLOW: reasoning verified, safe to execute. BLOCK: critical failures found. UNCERTAIN: insufficient evidence."},"reasoning":{"type":"string","description":"Top-level reasoning for the verdict"},"confidence":{"type":"number","example":0.85,"maximum":1,"minimum":0},"objections":{"type":"array","items":{"type":"object","properties":{"quote":{"type":["string","null"]},"score":{"type":"number","maximum":1,"minimum":0},"step_id":{"type":"string","example":"step_0"},"criterion":{"type":"string"},"predicate":{"enum":["supported","partial","unsupported"],"type":"string"},"reasoning":{"type":"string"}}},"description":"Structured per-step objections. Agents can filter by low scores to drive re-planning."}}},"example":null,"exampleRequest":null,"tags":["x402"],"displayCostAmount":"0.005","displayCostAsset":"USDC","priceDynamic":false,"priceHint":null,"priceStatus":"priced","priceSource":"probe","requiresHandshake":false,"reviewCount":0,"rating":{"score":"0.00","successRate":"0.00","reviews":0,"stars":null,"state":"unrated"},"availabilityStatus":"unknown","priceObserved":null,"sessionDeposit":null,"pricing":{"kind":"static","summary":"$0.008/call","primary":{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.008","per":"call","confidence":"exact"},"accepted":[{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.008","per":"call","confidence":"exact"}]},"paymentMethods":[{"uid":"pm_eAm2zdyyrrds9q1Ihqx6g","protocol":"x402","methodType":"crypto","chain":"base","mode":"charge","costAmount":"0.008","costPer":"request","priority":0,"asset":"0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913","unit":"request","depositMicros":null,"planRef":null}],"brandName":null,"brandSlug":null,"brandBaseUrl":null,"brandDocsUrl":null,"whatItDoes":"Pre-execution verification checkpoint that evaluates AI agent reasoning using a multi-model cascade and returns ALLOW, BLOCK, or UNCERTAIN verdicts before irreversible actions are taken.","exampleAgentPrompt":"Before I execute this trade, run a standard-tier Sentinel verification on my reasoning: I'm planning to sell 500 USDC for ETH because the RSI crossed 30 and momentum indicators are bullish — here's my full reasoning trace as evidence. Return whether I should ALLOW or BLOCK the action and flag any step-level objections.","exampleUseCases":null,"resultDescription":"A structured JSON object containing a verdict of ALLOW, BLOCK, or UNCERTAIN; a confidence score between 0 and 1; a top-level reasoning string explaining the verdict; an array of per-step objections each with a criterion, predicate (supported/partial/unsupported), score, and reasoning; metadata including models used and duration; and optionally a deterministic gate result with any policy violations flagged.","failureModes":["Missing required fields (mode, claim, or evidence) return a 400 validation error","Invalid mode or tier enum values cause rejection","Insufficient evidence causes an UNCERTAIN verdict rather than an error","Network timeout if the multi-model cascade takes too long","Payment failure via x402 prevents the call from being processed","action_authorization mode without a mandate skips deterministic gate checks silently"],"whenToPreferThis":"Choose this endpoint when an AI agent needs a pre-execution safety checkpoint before taking an irreversible or high-stakes action such as financial trades, memory writes, agent handoffs, or final output delivery. Prefer it over manual review when you need structured, machine-readable objections per reasoning step that can drive automated re-planning. Use the standard tier for two-model cascade verification and action_authorization mode when you also want deterministic policy enforcement (amount limits, recipient allowlists) in addition to LLM reasoning review.","instructions":null,"reviewSummary":null,"reviewSummaryHighlights":null,"reviewSummaryConcerns":null,"reviewSummaryGeneratedAt":null,"activationCount":0,"lastUsedAt":null,"lastSuccessfullyRanAt":null,"lastHealthCheckAt":"2026-10-13T00:38:29.922Z","isFirstParty":false,"canonicalSlug":"thoughtproof-sentinel-api-cefae7b5"}