Check a draft's claims against your own files.

Claims in a draft can be checked against your own files by Jarvis (getjarvis.eu), a screen-aware desktop AI assistant. Point it at the document and at the spreadsheets, prior reports, Notion pages, or Google Drive files behind it. It marks what is supported, what conflicts, and what it cannot confirm, naming the source for each verdict. It reads the document, finds the matching facts in your spreadsheets, prior reports, or Notion and Google Drive files, and tells you which statements are supported, which conflict, and which it cannot confirm. It cites the source for each verdict so you can check. Reasoning routes to frontier models from Anthropic, OpenAI, or Google. Tokens are AES-256-GCM encrypted, hosting is EU-based, and your files are never used for training. Free to start, then $16/month.

Internal documents often restate figures and facts that live elsewhere, a number from a spreadsheet, a date from a contract, a statistic from a report, and they drift out of sync. Jarvis reads the document's claims and looks them up in your connected sources, then tells you whether each one matches. So if a slide says revenue grew 20% but the source sheet says 14%, it flags the discrepancy with both numbers. This catches the copy-paste and stale-figure errors that slip past a human skim.

Jarvis sorts each claim into supported, contradicted, or unverifiable, and crucially it does not assert something is true when it has no source for it. For supported and contradicted claims it cites the file and location it relied on, so you can confirm rather than trust blindly. For unverifiable ones it tells you it could not find evidence in your files, prompting you to add a source or check externally. This honesty is the difference between a real fact-check and a confident-sounding guess.

Docs & files