Free · No signup · Real production engine

Audit your agent's
answer.

Paste an LLM answer and its sources. Get a numeric trust score in ~500ms.

Same deterministic engine Wauldo uses in production. No LLM-as-judge. Pure source-grounded fact-checking.

1. The claim (your LLM answer)

2. The source (your ground truth)

Mode:

No LLM-as-judge. No randomness. Pure deterministic source-grounded fact-checking.

STEP 1 · CLAIM EXTRACTION

Your text is split into atomic claims using sentence + clause boundaries.

STEP 2 · SOURCE MATCH

Each claim is matched against the source via token overlap, optionally BGE embeddings.

STEP 3 · CONTRADICTION DETECT

Numerical mismatches, negation conflicts, and opposition words override high overlap.

Want this in your stack?