Lenz

Lenz is an independent fact-checking API — a verification layer for AI-generated content. This connector brings it into Claude: check any factual claim, or re-check Claude's own output, against independent web sources.

Two tools, two speeds:

• Fast fact-check (assess_claim) — a 3-model verdict in ~5–10 seconds. The default for checking a claim.

• Deep fact-check (verify_claim) — the full pipeline: eight models across five stages (Framing → Research → Debate → Panel Review → Conclusion), ~90 seconds. It returns a sourced verdict — True, Mostly True, Mixed, Mostly False, or False — with an executive summary, the top sources, and the reasoning trace behind the result. Ask grounded follow-ups on any completed verification, and check remaining quota anytime.

Lenz grounds verdicts in external sources, independent of the context the model was given. It complements retrieval and groundedness checks rather than replacing them.

Lenz is unaffiliated with any model vendor. It is an outside check on AI output, not a model grading its own work.

The trace is the trust model. Every deep verdict includes the framing, sources, citations, debate, panel reasoning, and conclusion, so you can verify the verification. No verdict is infallible, which is why the trace ships with it.

Connect in one click with OAuth — no API key to paste — or with a free Lenz API key. Free plans include a monthly allotment of fast checks and a smaller number of deep verifications.