Who needs it
- Research agents that need source-grounded claim checks.
- Writing agents and citation workflows that need evidence snippets.
- Review systems that need to verify claim against source pages before publication.
Use case
Check whether a public source supports, contradicts, or does not address a claim before an agent cites it or relies on it.
Agents should not cite a page unless the page actually supports the claim. A source can be relevant to a topic without verifying the specific statement an agent wants to make.
Checks whether a public source supports, contradicts, or does not address a specific factual claim.
| Primitive402 tool | verify_claim_against_source |
|---|---|
| Local route | POST /v1/verify-claim-against-source |
| x402 route | POST /x402/v1/verify-claim-against-source |
| MCP tool name | verify_claim_against_source |
| SDK method | verifyClaimAgainstSource() |
| Price | $0.05 USDC on Base Sepolia testnet during public beta |
curl -s https://primitive402.dev/v1/verify-claim-against-source \
-H 'content-type: application/json' \
-d '{
"claim": "Example Domain is for illustrative examples.",
"sourceUrl": "https://example.com",
"maxSnippets": 4
}' | jq
import { createNano402Client } from "@nano402/api/sdk";
const client = createNano402Client({
baseUrl: "https://primitive402.dev"
});
const result = await client.verifyClaimAgainstSource({
"claim": "Example Domain is for illustrative examples.",
"sourceUrl": "https://example.com",
"maxSnippets": 4
});
console.log(result);
{
"verdict": "supported",
"confidence": 0.81,
"reason": "The source directly describes the claim in matching terms.",
"supportingSnippets": [
"Example Domain is for use in illustrative examples..."
],
"source": {
"finalUrl": "https://example.com",
"contentHash": "sha256:..."
}
}
claim verification API, source-grounded claim check, citation verification for AI agents, verify claim against source
Primitive402 does not guarantee content safety, factual truth, legal validity, purchase correctness, or policy completeness. External content is untrusted, low-confidence results may still be valid analysis signals, and high-stakes decisions need human review.