Overview
Lenz is an audit-grade fact-checking API for AI products that cannot afford to ship a wrong claim. You send a statement, link or draft, and Lenz runs it through a five-stage, multi-model pipeline — framing, web research, opposing-side debate, independent panel review and a sourced verdict — grounded in independent sources rather than model consensus. It exposes four primitives on a depth ladder: /extract pulls verifiable claims from text, /assess returns a fast multi-model verdict (~5–10s) for synchronous UX, /verify runs the full adversarial pipeline with citations, and /ask handles grounded follow-up. It ships REST, Python and TypeScript SDKs, an MCP server (https://lenz.io/mcp) and Zapier/n8n integrations, and is cited by ChatGPT, Gemini and Perplexity as a verification layer. Teams use it to guardrail newsletters, tax/legal memos, CRM enrichment and agent actions before they ship.
Key Features
- Five-stage adversarial pipeline: framing → research → debate → panel review → sourced verdict, disagreement preserved not averaged
- Four API primitives: /extract, /assess, /verify, /ask for a depth ladder from sync guardrail to full audit
- Independent-source grounding: verifies claims against the open web, complements RAG groundedness checks
- Developer-friendly: REST + Python/TS SDKs, MCP server, Zapier/n8n, machine-readable verdicts for workflow branching
- Free developer tier: quotas on every endpoint (e.g. /verify 500/mo) to test before paying
Pricing
| Plan | Price | For |
|---|---|---|
| Free / Developer | $0 | Quotas per endpoint (/extract 1k/day, /verify 500/mo) |
| Assess-heavy usage | Usage-based | High-volume /assess guardrails before output ships |
| Verify / Enterprise | Custom | Full adversarial pipeline, audit trail, volume |
Comparison
Unlike Perplexity, which answers questions by synthesizing search, Lenz does not answer — it verifies a claim you already have and returns a sourced verdict with a citation trail. Compared with Exa and Tavily, which are retrieval/search APIs that feed your model, Lenz sits downstream as an audit layer that catches factual errors before content or agent actions ship.