Lenz Fact-Check MCP Server

What is Lenz Fact-Check?

Lenz checks the factual claims in a draft or an answer against independent sources. Ask your assistant to check a claim or a whole text: Lenz finds the claims in it and returns a verdict and a confidence for each, and a deep check shows the sources behind a verdict.

How to use it

Add the server to your MCP client: https://lenz.io/mcp (Streamable HTTP). Sign in with your Lenz account when your client asks (OAuth, no API key needed); an API key from lenz.io/api-credentials also works as a Bearer header. Free accounts include monthly credits.

Then ask in plain language, for example: "Check the claims in this draft with Lenz."

Tools

• assess_claim: the quick check. A verdict (True, Mostly True, Mixed, Mostly False or False) and a confidence for each claim in a text, in about 15 seconds.
• verify_claim: the deep check for one claim. A 1-10 score, the key finding, warnings and the main sources with what each one says, in about a minute or two.
• get_verification: waits for a running deep check, or fetches a finished one.
• select_claims: picks which claims to check when a text holds several.
• ask_followup: answers a follow-up question about a finished deep check from its evidence.
• list_verifications: your recent deep checks.
• check_usage: credits left and the price of each tool.

How it works

Lenz researches each claim, has two models argue both sides from the sources, and has reviewers on models from different vendors score the evidence. Verdicts are directional, not absolute: the deep check shows its sources so you can judge for yourself.

Docs: https://lenz.io/integrations/mcp-server · Source: https://github.com/lenzhq/lenz-mcp · Privacy: https://lenz.io/privacy