What are you deciding?

A board of AI advisors from different companies reads your documents, answers on their own, debates once, and hands you a decision brief with every quote checked against the source. Keep the board as it is, or choose your own advisors.

Add options, what matters most, and context (optional, but it sharpens the advice)
Add documents (read in your browser, never uploaded to us)

Drop documents here or

PDF, Word, Excel, PowerPoint or text.

    The board · tap to seat or unseat

    Add an advisor

    Any model on OpenRouter that answers in text. Each request still asks for providers that don't keep or train on what they're sent; a model with no such provider will say so when it runs.

    Your OpenRouter key

    Verity runs on your own OpenRouter key, so the AI costs go on your OpenRouter account and your documents go only from this browser to the models.

    Tip: give the key a spending limit in OpenRouter. A decision usually costs between about $0.25 and $2.25, depending on the advisors.

    How it works

    1. Independent. Each advisor investigates alone, unaware the others exist, because knowing there's a room makes models posture or defer. Each reports its most consequential insight in about three sentences, with sources.
    2. Quote check. Every quote attributed to your documents is looked up in the text itself. Quotes that aren't there are flagged to the whole board.
    3. Conversation. Each advisor reads the others and advances the discussion: challenging assumptions, adding evidence, revising when persuaded. Agreement alone isn't progress.
    4. Decision brief. The advisors take turns writing it. It leads with the most consequential discovery, even if only one advisor found it, and keeps real disagreement visible. If the board isn't ready, the writer can ask for one more exchange. Underneath, each advisor leaves a one-line parting thought.

    A failure never reads as a success: if an advisor times out or its answer is cut off, the run says so and carries on with the rest. Long documents are read whole when they fit; when they don't, an advisor gets the sections most relevant to the decision, and the brief says so.

    Where your data goes