Minerval

Reads the page against the Minerval claim graph: flags misleading claims, explains contested ones, answers questions with citations.

Read the web with the claim graph switched on. Minerval maintains an open repository of the world's claims: what is asserted, what each assertion rests on, and how much the evidence supports it. This extension reads the page you're reading against that graph. WHAT IT DOES — Recognises the claims on a page and underlines each by what the graph knows: egregiously misleading claims in red, contested ones dotted, oversimplified and noteworthy ones more quietly. You choose how much markup you want; the default is conservative (only the worst). — Hover for the canonical claim, its graph status, and a one-line why. — Click for the full picture: how the claim decomposes into subclaims, the evidence and arguments on each side, and a link to the claim's public page, where you can challenge it or add evidence. — Ask questions in the popup chat and get answers grounded in the graph, with numbered citations to the claims they rest on — not vibes. Markup is non-destructive: annotations are drawn as an overlay and the page's text is never rewritten. PRIVACY, BY DESIGN Nothing is sent anywhere until you ask. Analyzing a page sends its readable text to the Minerval API, so analysis only runs when you press "Analyze page" — or on sites you explicitly opt into automatic analysis. Any site can be disabled entirely. No ads, no trackers, no analytics. SETUP Analysis runs against your Minerval account: sign in at minerval.ai, create an API key, and paste it into the extension's settings. Analysis and chat are metered against your account's monthly allowance; reading claim details is free. The graph is maintained by LLM administrators under a public constitution; every judgment carries a reasoning trace and is open to challenge. The project is open source: github.com/minerval-ai/minerval