Docs›Document review
…
Read with AI

Document review

Paste a draft or upload a PDF and Kvire returns peer-review-grade feedback — logical fallacies, weak arguments, suggestions — without you ever writing a prompt.

Why it exists

Researchers already use Claude or ChatGPT to critique their drafts. The friction: the prompts that actually produce useful feedback are long, specific, and tedious to write. “Find logical fallacies” as a one-liner gets you generic high-school-debate categories. The prompt that gets named fallacies anchored to specific sentences with suggested repairs is 200 words, and you don't want to type it every time.

Kvire ships those prompts. You click a lens, you get the output.

The lenses

  • General feedback — central claim, what's working, what needs work, top priorities.
  • Logical fallacies — named, anchored to specific quoted passages, with repair suggestions.
  • Weak arguments — load-bearing claims that don't hold up, ranked by importance.
  • Best counter-arguments — steelmanned objections from a fair-minded opposing reader.
  • Concrete suggestions — structural, argumentative, and prose-level edits with specific quotes.

How it works

  1. Open /review.
  2. Paste text (≥150 words) or upload a PDF (≤25 MB). Stored on your account.
  3. Click any lens — Claude streams the response in.
  4. Run as many lenses as you want. Results are cached, so re-opening costs nothing.

Cost

Reviews use Anthropic prompt caching: the document is sent once with cache_control: ephemeral, then each lens prompt runs against the cached context. Cached input tokens cost ~10% of full input, so reviewing a 6000-word paper with 5 lenses costs ~$0.05 instead of ~$0.30 on the underlying API.

What it's not

  • Not a copy-editor. Sentence-level grammar feedback isn't the goal — Grammarly is good at that.
  • Not editorial. The output is AI-generated; treat it as a peer-review pass, not a final judgment.
  • Not for non-textual content. PDFs that are scanned images can't be extracted.

Roadmap

Next planned addition: a missing-citations lens that uses Kvire's catalog to suggest specific papers your draft should cite — something neither ChatGPT nor Grammarly can do. Track progress on the public roadmap.