Quickstart
Get up and running with Check in under 5 minutes.
Quickstart
This guide will help you verify your first piece of AI-generated content using the Check API.
Prerequisites
Before you begin, make sure you have:
- A Check account - Sign up here if you don't have one
- An API key - Generate one from your Dashboard
Step 1: Get Your API Key
Navigate to Dashboard → API Keys and click "Create New Key". Give it a descriptive name and save the key securely - you won't be able to see it again!
Keep it secret
Your API key grants access to your account. Never expose it in client-side code or public repositories.
Step 2: Install the SDK (Optional)
While you can use the REST API directly, our SDKs make integration easier:
Step 3: Make Your First Request
Use the following code to verify a piece of content:
Step 4: Understand the Response
A successful response looks like this:
{
"id": "ver_abc123",
"status": "completed",
"verdict": "true",
"confidence": 0.95,
"decision": "accept",
"reasoning": "The claim is factually accurate. The Eiffel Tower is indeed located in Paris, France, and construction was completed in 1889 for the World's Fair.",
"paradigmResults": [
{
"paradigm": "reasoning",
"verdict": "true",
"confidence": 0.94,
"reasoning": "The claim aligns with established historical facts"
},
{
"paradigm": "tool",
"verdict": "true",
"confidence": 0.97,
"reasoning": "Confirmed via multiple authoritative sources"
}
],
"processingTimeMs": 1850,
"createdAt": "2024-01-15T10:30:00Z",
"completedAt": "2024-01-15T10:30:02Z"
}Response Fields
| Field | Description |
|---|---|
id | Unique verification identifier (prefixed with ver_) |
status | pending, processing, completed, or escalated |
verdict | true, false, or uncertain |
confidence | Confidence score from 0 to 1 |
decision | Recommended action: accept, refine, escalate, or reject |
reasoning | Human-readable explanation of the verdict |
paradigmResults | Individual results from each verification method |
processingTimeMs | Total processing time in milliseconds |
Understanding Decisions
| Decision | Confidence | Condition |
|---|---|---|
accept | >= 0.95 | High confidence - safe to use |
refine | 0.70 – 0.95 | Moderate confidence - may need clarification |
escalate | < 0.70 | Low confidence + paradigm disagreement - flag for human review |
reject | < 0.70 | Low confidence + paradigm agreement - do not use |
Step 5: Choose Your Verification Methods
Check offers 7 verification methods with different accuracy/speed tradeoffs:
| Method | FPR | Speed | Best For |
|---|---|---|---|
formal | 0% | Fast | Mathematical claims |
tool | ~1% | Slow | Real-time fact-checking |
reasoning | ~4-5% | Medium | General fact verification |
biprm | ~5-7% | Medium | Complex reasoning chains |
entropy | ~8-10% | Medium | Consistency validation |
ensemble | ~15-20% | Slow | High-stakes decisions |
semantic | ~20-30% | Fast | Quick similarity checking |
Configure methods with weights (0.0 to 1.0) based on your needs:
// Fast verification for chatbots
const result = await client.verifyAndWait({
content: 'Your claim here',
methods: { reasoning: 1.0 }
});
// High-accuracy verification for important content
const result = await client.verifyAndWait({
content: 'Your claim here',
methods: {
reasoning: 1.0,
tool: 1.0,
ensemble: 0.5
}
});
// Mathematical claims
const result = await client.verifyAndWait({
content: '2 + 2 = 4',
methods: { formal: 1.0 }
});Next Steps
- Learn about Authentication for production use
- Explore the full API Reference
- Set up Webhooks for async notifications
- Try Batch Processing for bulk verification
- Read about Verification Methods in depth
- See Rate Limiting for usage limits