SourceScore

5 minutes · no signup required

Find your first catalog match

Three calls. Pick the language you're already in. The public API is free and needs no account or key. Standard network abuse controls may apply.

Step 1 — Find a candidate record

Send a free-form statement; get back candidate catalog records with similarity metadata and canonical citation URLs. Compare the returned statement and cited evidence with your input.

curl

curl -X POST https://sourcescore.org/api/v1/verify \
  -H "content-type: application/json" \
  -d '{"claim":"The Transformer architecture was introduced in 2017."}'

JavaScript / TypeScript

const r = await fetch("https://sourcescore.org/api/v1/verify", {
  method: "POST",
  headers: { "content-type": "application/json" },
  body: JSON.stringify({
    claim: "The Transformer architecture was introduced in 2017.",
  }),
});
const { bestMatch } = await r.json();
console.log("Candidate record:", bestMatch?.statement);
// → "Transformer architecture introduced_in_paper: Attention Is All You Need (Vaswani et al., 2017)." 1.0

Python

import requests
r = requests.post(
    "https://sourcescore.org/api/v1/verify",
    json={"claim": "The Transformer architecture was introduced in 2017."},
)
best = r.json().get("bestMatch")
if best:
    print("Candidate record:", best["statement"])
    print(f"Citation: https://sourcescore.org/claims/{best['id']}/")

What you get: a typed candidate record with its statement, legacy editorial-confidence metadata, claim ID, and a canonical URL. A match is retrieval—not proof that your input is true.

Step 2 — Search or browse

When the user's intent is browse-mode rather than verify-mode, use search to surface the top-K relevant claims.

# Search
curl "https://sourcescore.org/api/v1/search?q=context+window&limit=5"

# Full catalog (one call, no pagination at v0)
curl https://sourcescore.org/api/v1/claims.json | jq '.claims | length'
# → 100

Step 3 — Fetch the full record

When you need the full provenance (verbatim source excerpts, SourceScore-issued HMAC integrity metadata, methodology version) for a specific claim, fetch the envelope.

curl https://sourcescore.org/api/v1/claims/<claim_id>.json

Replace <claim_id> with the id returned from step 1 or step 2. The envelope ships with signature HMAC metadata is not publicly independently verifiable because the shared secret is not published. Refetch this canonical URL and inspect the cited evidence.

That's it.

You just made a catalog-matching API call. Wire this into your LangChain, LlamaIndex, or tool-call chain via the integration guides. Or pin a specific high-confidence claim into your prompt as context. If you need higher volume, private claim sets, or a service commitment, request higher-volume access.

What now?