OpenAI's Computer History Turns Desktop Activity Into Context for ChatGPT and Codex
OpenAI has introduced Computer History, an opt-in feature in the ChatGPT macOS desktop app that records user activity across applications and websites and converts it into a searchable timeline that ChatGPT and Codex can reference when responding to requests. The feature was announced in the ChatGPT desktop app changelog on August 13, 2026, and is currently available to ChatGPT Pro, Business, and Enterprise customers on macOS.
Computer History captures what OpenAI calls "events" rather than raw screen captures. Unlike Microsoft's Windows Recall, which relied heavily on screenshots, Computer History does not capture images, videos, or audio. Instead, it logs discrete user actions — clicks, keystrokes, and application-level interactions — and builds a timeline from them. That timeline then serves as contextual memory: when a user issues a prompt, ChatGPT or Codex can draw on recent activity to infer intent, suggest automations, or pick up tasks left half-finished.
A demo video by Dominik Kundel of OpenAI's Developer Experiences team illustrates the workflow. In the recording, the app identifies the last document Kundel edited, checks whether it was shared with colleagues via Slack, and delivers a recap of how he spent his morning. The use case is less about transcription and more about giving the model a persistent, structured understanding of what the user has been doing across the operating system, not just within the ChatGPT window.
Computer History is off by default for Pro, Business, and Enterprise users. Pro users can choose to enable it. Business and Enterprise administrators can manage the feature for their organizations. Individual users can exclude specific apps and websites from tracking and delete individual entries from the timeline. Ari Weinstein, Product and Engineering manager at OpenAI, stated on X that the feature will automatically ignore content in incognito or private browser tabs.
OpenAI's official X account described the goal in straightforward terms: with Computer History, "future interactions feel more personalized and require less explanation." The feature brings context from selected apps and websites into ChatGPT and Codex, turning activity into memories and a timeline that persists across sessions.
The architectural distinction from Windows Recall matters beyond messaging. Recall stored periodic screenshots locally and exposed them through a searchable interface, which drew intense scrutiny over the risk of sensitive information being captured and later accessed. By logging structured events rather than pixel data, Computer History reduces the surface area for certain classes of leakage — you cannot extract a screenshot of a banking session from an event log. The trade-off is that structured event data is harder for users to audit visually. A screenshot is immediately legible; a list of keystroke and click events requires tooling to interpret, which places greater weight on the integrity of OpenAI's own processing and deletion controls.
The opt-in framing is consistent with how OpenAI has handled its most privacy-sensitive features, and the granular exclusions — per-app, per-site, per-entry — give technically literate users the controls they would expect. The admin-level management for Business and Enterprise tiers aligns with what IT teams responsible for fleet-wide deployments would need before enabling something that broadly captures workstation activity.
Looking at the broader trajectory, Computer History sits at the intersection of two trends that have been converging for the past two years: the expansion of context windows into effectively unbounded territory, and the push toward agent-like workflows where models act on a user's behalf across multiple tools. A model that can remember what you did this morning is a prerequisite for one that can take over where you left off. Whether that capability is net positive depends entirely on execution and trust. The technical approach — structured events, no screenshots, incognito awareness, granular controls — is a reasonable design for the problem. The remaining question, as with any feature that logs user behavior at this granularity, is what happens to that data downstream: how long it persists, whether it feeds model training, and what guarantees users have that deletion is irreversible. OpenAI says user actions are turned into training data. The controls are present. The burden of living up to them is ongoing.
