CellSource

The case for CellSource is narrow and clear: you work in Excel on a Mac, and you need each figure traceable to a source document. On that job it does the right things. Evidence lives in cell metadata, so re-sorting or pivoting the sheet doesn't shred your links — a genuine failure mode with shape-based annotation tools. Clicking a linked cell gives you the snippet, page number, and a zoom to the match. We'd reach for it during invoice and receipt reconciliation, or when assembling a SOC 2 evidence package where a reviewer needs to verify a figure in one click. The built-in OCR pulling text from handwritten invoices and scanned receipts matters here, because that's exactly the material that usually resists search. Pick it if the alternative is a Windows VM. The vendor's whole pitch is the DataSnipper alternative for Mac, and cross-platform Office.js support is the practical difference — no Citrix, no Parallels, no lag. Processing stays local in the browser; the vendor says it does not store or see your PDFs, which is the thing that makes confidentiality review easier to sign off on. Pass on it if your need is insight rather than verification. There is no semantic search, no summarisation, and no Q&A across a document set — it won't answer "what's the trend in our vendor spend" the way a general-purpose AI research tool will. Search is literal: amount, invoice number, keyword. Caveats worth weighing before you commit. The product is a public beta, so budget for rough edges. On the current profile it's English-only, and while the vendor lists iCloud Drive, Dropbox, and local file access, the homepage publishes no integrations page, documentation hub, or changelog — so integration depth beyond that is unverified. Batch upload exists per the profile, but the vendor page