Privacy & data handling
Last updated: August 2026
How your library, AI requests, and billing information are handled.
How data moves
Your library stays local. AI features send needed context through the backend to models, including automatically selected context and manually attached passages.
01
Local PDF
Imported files stay on your computer.
02
Local index
Searchable passages are built on-device.
03
Context for the task
Needed manuscript context, matched passages, and manually attached text or figures may leave the device.
04
Flowing backend
Requests are proxied over HTTPS. You do not add your own API keys.
05
Model providers
DeepSeek for keyword recall tasks; Kimi (Moonshot) for Ask, polish, and figures.
06
Result returned
The suggestion comes back to the desktop app, with evidence cards when applicable.
Published retention commitments for request bodies and third-party model providers are not yet available. Labs and institutions should contact [email protected] before using Flowing with restricted or confidential material.
Your PDF library stays local
Reference PDFs you import are stored and indexed on your device. Flowing parses text locally so you can search and attach passages without uploading your entire library to the cloud by default.
AI features send the manuscript context and matched library passages needed for the task, plus any content you attach, to Flowing’s backend—not the full PDF library.
Cloud AI models
Flowing routes AI requests through its backend to third-party models. Automatic snippet recall and related retrieval tasks use DeepSeek; Ask with context and evidence-grounded revision use Kimi (Moonshot), including multimodal understanding when figures are cited.
You do not configure API keys yourself. Default model versions may change with releases. Pro+ subscribers can access stronger model options.
What leaves your device
AI requests can include prompts, manuscript context and matching passages selected by the feature, and snippets you attach manually. Attached figures may be sent for model understanding; transmitted context is not limited to manually selected excerpts.
Your full PDF library, local search index, and draft files are not uploaded by default.
Web paper recommendations
Auto-recommend papers from the web is off by default. The first time you turn it on, you confirm a privacy notice. Keywords are sent to Flowing’s recommendation service; paragraph body text is sent only if you opt in.
OpenAlex metadata is used to verify that recommended works exist. Links are not invented.
Deletion and institutional use
Local PDFs, drafts, and indexes can be removed on your machine. To ask us to delete server-held account or billing records, email [email protected]. Stripe retains payment records under its own policies.
A DPA and full sub-processor retention schedule are not yet published. Research groups that require written compliance terms should contact [email protected] before using Flowing with restricted or confidential material.
Billing and account data
Paid plans and prepaid token packs are processed through Stripe. Checkout uses claim codes generated in the desktop app. We store only what is needed to verify entitlements and complete transactions.
For billing questions, contact [email protected].
Website analytics
This marketing site may use Google Analytics when enabled. The desktop application’s usage metering for AI features is separate from website analytics.