# Show HN: Drifty – AI Focus agent shuts down distractions tabs while you work > Source: > Published: 2026-06-30 03:18:25+00:00 **Live** **Visual Studio Code** Landing page edits Focus **Slack** Context switch Drift ### Automatic capture Records your active app, site, and session in the background. No timer to start, no manual tagging to clean up later. Runs in the backgroundWhy drifty drifty records the apps, sites, and sessions behind your day, then classifies the work context without manual timers or tags. Records your active app, site, and session in the background. No timer to start, no manual tagging to clean up later. Runs in the backgroundUses fast model routes to label captured activity as Focus, Neutral, or Drift without asking you to tag it. Gemini, DeepSeek, QwenNo screen recording or saved screenshots. Run AI locally, in drifty Cloud, or with your own key. No screen recordingAI CLASSIFICATION drifty reads activity context and classifies each moment as focus or drift. AI MODES The same activity classification can run in drifty Cloud, through your own API key, or on your Mac. High-performance AI classification, managed by drifty. Use your own [OpenRouter](https://openrouter.ai/) key and control usage. Run classification directly on your Mac. 01Track & visualize drifty records your apps, sites, and sessions in three-minute blocks — no manual logging, no timers to start. drifty records the app, site, and session while you work. By the end of the day, the timeline is ready to review. Each block is three minutes, labeled as focus, neutral, or drift. Zoom out for the full day, or switch views to see what filled each hour. Captured segments stack into a weekly view automatically. Each card keeps its source apps and sites, with daily rings and weekly bars for a faster summary. 02AI classification Categories show the shape of the day, apps and sites show where it happened, and activity details show what the time was actually spent on. 03Insight See when focus sessions are most common, when drift increases, and which patterns repeat across the week. See when focus holds, when drift appears, and how your day changes hour by hour. A weekly heatmap shows when focused work happens most often. The profile summarizes the pattern for planning. 04Drift Alarm When a drift starts, drifty gives you the next move in context: return to a recent focus tab, take a short break, or mark the moment as not a drift. The alarm keeps your recent focus surfaces one click away, so the recovery path is visible before the drift turns into a long session. For stricter focus sessions, drifty can close the distracting tab and switch directly to the most recent focus tab. Comparison A focused comparison of automatic tracking, same-app context, AI classification, and control. | drifty | Rize | RescueTime | Timely | | |---|---|---|---|---| | Automatic tracking | Supported | Supported | Supported | Supported | | Same-app understanding | Supported | Partial | Not supported | Partial | | Focus / Drift classification | Supported | Partial | Partial | Not supported | | AI control | Supported | Partial | Not supported | Not supported | | Starting price | From $0/mo | From $12.99/mo | From $9/mo | From $11/mo | No. drifty runs quietly on your Mac, records app, site, and session context in the background, and turns the day into a timeline you can review later. drifty looks at the activity context around each session: the app, the site, the title, and how that work fits into the rest of your day. It is meant to understand the work, not just the app name. No. drifty is not a screen recorder and does not save screenshots. The product is built around activity context, not visual surveillance. You choose. Classification can run through drifty Cloud, on your Mac with Local AI, or with your own [OpenRouter](https://openrouter.ai/) key. You can switch modes later. Activity capture stays local on your Mac. Cloud and OpenRouter classification need a connection; when you are back online, drifty can catch up and classify the captured sessions together. Local AI can classify on-device after you set up a local model.