OpenAI launched Chronicle in Codex — a Windows Recall equivalent for developers

OpenAI has released the Chronicle feature in the Codex agent application — a tool that periodically takes screenshots of a developer's screen so that AI can better understand the context of what's happening. Chronicle is operating in preview mode and is designed to provide Codex with more information about the user's working environment. Screenshots allow the agent to see what code is open in the editor, what errors are displayed in the terminal, what documents the developer is studying — all of this helps form more accurate answers and suggestions.

The feature directly echoes Windows Recall from Microsoft — a tool that also makes regular screenshots of user activity to create a "memory" of the system. Recall triggered a wave of criticism after its announcement in 2024: experts pointed to privacy risks, the possibility of password leaks, banking data and personal correspondence captured in the frame. Microsoft postponed the full launch several times and seriously reworked the data storage mechanism.

OpenAI, it seems, is banking on the fact that in a professional context — among developers working with code — such an approach will be received more positively. The target audience for Chronicle is more uniform: a programmer's screen contains mainly code, terminal, and documentation, rather than personal correspondence or financial services. Nevertheless, questions about where screenshots are stored, who has access to them, and how long they are preserved remain without a detailed public answer.

Codex is already positioned as an autonomous agent capable of performing multi-step tasks in a code base without constant human involvement. Chronicle logically complements this concept: an agent that "sees" the screen can make more meaningful decisions and less often ask the user to clarify the context manually. The release of Chronicle in preview shows that the race for "screen memory" in AI tools is gaining momentum.

After Microsoft and now OpenAI, this approach will likely appear in other market players — the question is only how transparent the conditions for working with the collected data will be.