API

Build AI you can trust into anything.

One API call. Full confidence analysis, hallucination detection, hedging analysis, entity verification, and improvement suggestions. Add it to your chatbot, your content pipeline, your code review system, or anything that touches AI output.

Verify chatbot responses

Score every response your customer-facing bot generates before it reaches the user.

Audit AI-generated reports

Catch hallucinated statistics, fabricated citations, and unsourced claims in automated content.

Check AI-written code

Flag deprecated APIs, security patterns, and anti-patterns in AI-generated code before it ships.

Grade LLM outputs

Benchmark and compare model quality across providers with structured scoring on 7 axes.

Quick start

curl

curl -X POST https://consensus-production-6eeb.up.railway.app/api/v1/check \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Your AI response text here",
    "query": "The original question",
    "tier": "protocol"
  }'

No API key required for testing. Production keys available with Full Stack ($19/mo) or BYOK ($9/mo) plans.

Try it live

The example below is pre-loaded with a classic AI hallucination. Hit send to see what ASURIQ catches.

Try it live

Response schema

confidence.scorenumberOverall confidence (0–1). Factors in domain risk, complexity, hedging, source quality.

confidence.axes[]array7-axis breakdown: factual, completeness, recency, balance, domain, source, coherence. Each scored 0–1.

confidence.bottleneckstringPlain-English explanation of what’s dragging the score down.

hallucinationobjectHallucination risk score (0–1), assessment, signals array, fabricated entity count, uncited claim count.

hedgingobjectCount and assessment of hedging phrases. Catches when AI sounds certain but qualifies every claim.

toneobjectPrimary tone (assertive/hedging/reassuring/academic/cautionary/conversational) + domain mismatch flags.

fmeaobjectError visibility score. How likely would the user notice if this answer was wrong?

entities[]arrayExtracted studies, organizations, statistics. Each flagged as verifiable with type classification.

citationsobjectClaim count vs. cited count. Lists specific uncited claims with strong assertions.

contradictions[]arrayDetected internal contradictions with the specific conflicting sentences.

improvementHintsobjectPrimary hint + optional rewritten prompt. Generated from the weakest confidence axis.

keyClaimobjectThe single most important sentence, ranked by salience scoring.

What each tier returns

Pass tier in your request to control analysis depth.

freeconfidence.score, mode, classificationBadge data

t1+ hedging, tone, keyClaimFull checks

t2 / protocol+ 7-axis, hallucination, FMEA, entities, citations, contradictions, improvementHintsEverything

API pricing

Full Stack

$19/mo + $0.03/query

We handle the LLM routing. Send text, get the full analysis. No API keys to manage.

BYOK

$9/mo + $0.03/query

Bring your own API keys. Same verification engine, your models, lower cost. You pay providers directly.

Every AI output deserves verification.

Add hallucination detection, confidence scoring, and entity verification to your product in one API call.

Start building

Questions? api@asuriq.dev