Human review and validation rules for AI you can trust

Catch compliance issues, update rules instantly, no deployments needed

Catch Privacy leaks Non-compliance Policy issues Dubious links PII exposure before they become incidents

Fix JSON problems HTML problems Markdown fails Format errors Invalid syntax Missing fields with human oversight

Without Check Your AI

  • Bad format: JSON/Markdown/HTML errors
  • Missing required fields or content
  • Non-compliant content, PII, or bad links
  • No audit trail for compliance
  • Hardcoded rules require deployments

Hours of debugging

Multiple expensive retries

With Check Your AI

  • Rules catch any issue instantly, approve, reject or send for review
  • 👤 Human fixes format, content, or compliance
  • Corrected response ready to use
  • Webhook sends to your app

Fixed in seconds

Update rules without deploying

No credit card required Talk to founders about pricing

Enterprise-ready
Built with Elixir
Fast API Response

Why I built this

I was using AI to generate document summaries. But the AI kept getting things wrong, missing titles, incorrect structure. I'd have to retry, sometimes 3+ times until the AI finally got it right.

I started writing custom validation rules to reject bad summaries automatically. But hardcoding these rules meant every change required a deployment. I needed flexibility.

And retrying wasn't just annoying, it was expensive. Multiple API calls for the same content. For format errors, a human could fix it in 10 seconds. Sometimes you just need human judgment.

That's when I built Check Your AI.

Lua rules catch issues automatically,format errors, missing content, compliance problems etc. I can update validation logic instantly without deploying. A human reviewer fixes issues in seconds. The corrected response goes to my webhook. I get it right the first time, and I can iterate on my validation rules without redeploying.

The Problem Retrying Can't Solve

Some Issues Need Human Judgment

Retrying the same AI call won't fix compliance violations, PII leaks, or policy issues, you need predictable validation rules and human oversight

Retrying Can't Fix This

  • PII Detection

    AI might leak names, emails, SSNs again on retry

  • Content Policy

    Same problematic output, different attempt

  • Regulatory Audit Trail

    No record of who reviewed or why it was rejected

  • Context-Dependent Judgment

    AI can't understand nuanced compliance requirements

  • Unpredictable AI Behavior

    Each retry might produce different results no consistency

Human Review Catches It

  • PII Detection & Removal

    Humans spot sensitive data AI models miss

  • Policy Enforcement

    Reviewers ensure outputs meet your guidelines

  • Audit Trail

    Every review logged for compliance reporting

  • Contextual Judgment

    Humans understand nuanced requirements AI can't

  • Predictable Rules, Predictable Outcomes

    Lua validation rules are deterministic, no guessing, no surprises

Industries Where Compliance Isn't Optional

Healthcare

HIPAA compliance requires human oversight

  • Remove patient names, DOBs, medical IDs
  • Verify medical summaries for accuracy
  • Audit trail for compliance officers

Finance

SOC 2 / audit requirements need review logs

  • Verify financial advice accuracy
  • Catch potential regulatory violations
  • Document human oversight for auditors

Legal

Privileged information can't leak

  • Redact confidential client details
  • Ensure contract summaries are accurate
  • Human-in-loop for attorney-client privilege

Retrying Won't Protect You

If you're in healthcare, finance, legal, or any regulated industry human review isn't optional. Check Your AI gives you the compliance infrastructure you need.

Is Check Your AI Right For You?

Calculate your tier and potential savings

💰 Cost Comparison

Without Platform: $197.13
Base: $118.75 + Retries: $78.38
With Platform: $268.74
Base: $118.75 + Fee: $149.99
⚠️ $71.62/mo cheaper to retry

At your 22% retry rate, retrying is more cost-effective. However, retrying won't work in all cases—sometimes the LLM simply can't produce the right output. Consider our platform for:

  • ✓ Better output quality (human reviewers catch what AI misses)
  • ✓ Instant rule updates (no deployments)
  • ✓ Compliance requirements (audit trails, human oversight)
  • ✓ Predictable validation (Lua rules vs unpredictable retries)
10,000 reviews/month
100 1K 10K 50K 100K

How often do you retry AI calls due to errors?

22%
Best: 19% | Typical: 22% | High: 35%+
19% 25% 30% 35% 40% 50%
📊 Startup Tier

$149.99/month

Your volume: 10,000 reviews/month

✓ Perfect fit!

* Calculations based on 2,500 tokens per call and 3 retries per failure. Default 22% retry rate based on real-world benchmarks (SWE-bench Verified): Claude Opus 4.5 (19%), GPT-5 (25%), Gemini 3 Pro (24%). Actual results vary based on your specific setup.

The real cost of API retries

Most developers don't realize how much they're spending on retries

Scenario Without Check Your AI With Check Your AI Result
Malformed JSON/Markdown/HTML Multiple expensive retries One quick human fix ~10x cheaper
Missing required fields/content 2-3 API retries Fixed in seconds Way cheaper
PII or non-compliant content Can't retry need human judgment Human review catches it Prevents issues

Why pay for the same content multiple times?

Old Way

Retry approach

Burning money on failed attempts

Multiple API calls for the same response. Each retry costs the same as the original call.

1.6x the original cost
OR
Smart Way

Review approach

Fraction of the cost

Human fixes it right the first time. No expensive retries, just intelligent correction.

-44% cost savings

Want to know how much you could save?
Talk to our founders about your use case. We'll figure out the right pricing together.

Integration in minutes

How It Works

Simple API integration, powerful compliance

1

Your app calls AI API

Response has issues: bad JSON/Markdown/HTML, missing fields, PII, non-compliant content

❌ Multiple issues: Parse error, missing required field, contains PII
2

POST to Check Your AI

Lua rule instantly evaluates the response and routes it

Valid responses → Auto-approve
Low quality → Auto-reject
Needs fixing → Human review
3

Human fixes in seconds

Reviewer corrects format, adds missing content, removes PII way cheaper than retrying

Average review time: 10-30 seconds

Webhook to your app

Get corrected response instantly. No retry needed, compliance maintained

✓ Error correction 🛡️ Compliant by design ⚡ Instant rule updates 📋 Full audit trail
See it in action

Watch How It Works

See the complete review flow in 2 minutes

API → Review → Webhook flow
Lua rules in action
Review interface walkthrough

Auto-Rules with Lua

Write custom validation rules or use built-in templates

validation_rule.lua
-- Reject low-confidence responses automatically
if confidence and confidence < 0.2 then
return 'reject'
end
-- Auto-approve valid JSON, route errors to human review
if output and is_valid_json(output) then
return 'approve'
end
return 'review'
-- Or check specific fields
if output.status == 'success' and output.data then
return 'approve'
end
return 'review'
-- Require review if contains PII
if has_pii(output) then
return 'review'
end
Use built-in templates or write your own
Auto-approve valid responses (free, instant)
Auto-reject low-quality responses (save review costs)
Route errors to human review automatically

Simple, transparent approach

Let's talk about your use case and figure out pricing together

Talk to our founders

We're figuring out pricing by talking to developers like you. Let's discuss your use case and find the right fit together.

🚀 Early Access Pricing - Lock in this rate forever

Choose your plan

Start with self-service or scale with our managed review service

Developer

For solo developers

$14.99 /month

You review your own content

  • Up to 1,000 reviews/month
  • API + webhooks + Lua automation
  • Basic auto-rules (3 rules max)
  • 12 pre-built templates
  • Email support
  • 1 user
MOST POPULAR

Startup

For small teams

$149.99 /month

Your team reviews your content

  • Everything in Developer, plus:
  • Up to 10,000 reviews/month
  • Team collaboration (up to 5 users)
  • Unlimited Lua auto-rules
  • Higher API limits (500 req/5min)
  • Priority support
  • 30-day audit logs

Enterprise

For high volume & compliance

Custom

We can handle reviews for you

  • Everything in Startup, plus:
  • Unlimited reviews
  • We handle reviews for you
  • Unlimited team members
  • Custom API limits & SLAs
  • Dedicated support
  • Coming: SSO, SOC 2, data residency

All plans include: Lua scripting, webhooks, API access, audit logging

Need higher volume or custom features?

Built for AI outputs that matter

Quality assurance for critical AI workflows

Customer Chatbots

Prevent embarrassing AI mistakes before customers see them

Marketing Copy

Maintain brand voice and tone across all AI-generated content

Support Automation

Accuracy is critical when AI handles customer support

Compliance Docs

Meet regulatory requirements with human oversight

Frequently Asked Questions

Everything you need to know about Check Your AI

How does Check Your AI work?

Instead of sending AI outputs directly to your application, you send them to Check Your AI first. Our platform runs your Lua validation rules to auto-approve, auto-reject, or route to human review. On Developer and Startup tiers, you or your team handles reviews. On Enterprise tier, we provide managed review services. Either way, the corrected response goes to your webhook no retry loop needed.

What is Lua scripting and why do I need it?

Lua is a simple scripting language that lets you write validation rules without deploying code. You can check for missing fields, validate JSON structure, detect PII patterns, enforce content policies, and more. We provide custom helper functions to extend Lua's native capabilities (like `is_valid_json()`, `has_pii()`, etc.), and we can build new helper functions on request based on your needs. Rules are deterministic (same input = same result) and can be updated instantly without waiting on engineering deployments.

How much does it cost?

We're in early access and figuring out pricing by talking to developers. We have tiered plans starting at $14.99/month for up to 1,000 reviews. For higher volumes or custom needs, we work with you to find the right fit. Email us at hello@checkyour.ai to discuss your use case.

How fast are human reviews?

It depends on complexity. Simple fixes (like correcting JSON formatting) take seconds. Complex compliance reviews requiring deep analysis could take minutes to hours. On Developer and Startup tiers, you control review speed since your team handles it. On Enterprise tier with managed reviews, we work with you to set SLAs based on your needs.

What AI models and providers do you support?

Check Your AI is model-agnostic. We work with any AI provider (OpenAI, Anthropic, Google, open-source models, etc.) and any output format (JSON, text, HTML, markdown, etc.). You send us the AI output, we validate it, and send back the corrected version. It doesn't matter which model generated it.

Is my data secure?

Yes. All data is encrypted in transit and at rest. Our reviewers sign NDAs and are trained on data handling policies. We're building toward SOC 2 compliance and can work with you on specific security requirements. For sensitive data, we also support on-premise deployments (Enterprise tier).

Can I use my own reviewers?

Yes! On Developer and Startup tiers, you handle your own reviews—invite your team members or domain experts to access the review dashboard. On Enterprise tier, we provide managed review services, but you can also bring your own reviewers if you prefer.

What happens if I exceed my tier's review limit?

All reviews count toward your tier limit—both automated (Lua rules) and human reviews. If you exceed your tier's monthly review quota, you can either upgrade to the next tier or purchase additional review capacity. We'll work with you to find a solution that fits your needs.

Ready to catch what AI can't?

Build compliant AI with human oversight

Free to start • No credit card required • Talk to founders about pricing