Private by architecture.

No analytics SDKs, no crash reports, no ad tracking. Fylow talks to our servers only to do the work — sorting, chat, your license, update checks — and usage is metered for billing as numbers, never content.

Content Access builds a search index of your documents and screenshots — a single file on your Mac, never uploaded, never synced. Turn it off and the index locks shut.

Ask a question, and one file travels — the best candidate, only the excerpt needed. Run a folder sort, and it reads that folder within hard per-run limits, because that’s the job. Background reading doesn’t exist here.

Your text is processed in memory — never stored, never trained on. Our server checks the provider’s retention flag on every model response and refuses when it’s missing; voice transcription is covered by the same contract.

Your conversation history lives in files on your Mac — not in anyone’s database. Each message is processed in memory on its way to the model, then forgotten.

Before a move, a copy lands in a local vault (very large files are tracked by location). Moves are undoable from history while the vault holds the copy — 7 days by default, yours to extend — and that safety net never leaves your machine.