What it does Dayflow runs quietly on macOS, captures lightweight screen chunks, and uses your choice of AI provider to turn raw activity into a chronological timeline of what you actually did. It generates daily standup summaries, weekly analytics, distraction tracking, and lets you chat with your own work history in natural language. Everything stores locally by default in ~/Library/Application Support/Dayflow/.

The interesting bit The privacy architecture is the real design decision here. You can run entirely on local models via Ollama or LM Studio, or selectively use cloud providers (Gemini, ChatGPT, Claude) depending on whether you value privacy, cost, or summary quality for a given session. Most time trackers log window titles; Dayflow tries to understand context — distinguishing “two hours in Cursor” as shipping a feature versus debugging auth versus lost in setup.

Key highlights

• Automatic timeline generation from screen activity, no timers or manual notes
• Context-aware summaries that interpret on-screen content, not just app names
• Daily standup prep with yesterday’s highlights, today’s priorities, and blockers
• Natural-language chat interface for querying your own work history
• Weekly review with focus patterns, app usage breakdowns, and distraction flags
• Markdown export for any date range
• Configurable automatic cleanup with storage limits
• macOS 14+ only; install via DMG or brew install --cask dayflow

Caveats

• Requires macOS Screen & System Audio Recording permission, which is a significant trust boundary
• Cloud AI providers receive activity data needed for analysis; local models avoid this but may trade quality
• Swift/Xcode build from source is straightforward but not cross-platform

Verdict Grab it if you’re a Mac user who wants automated work journaling without feeding another SaaS your screen history. Skip it if you’re on Linux/Windows, or if the idea of any screen recording — even local-first, open-source — feels inherently creepy regardless of privacy promises.