Codex Chronicle Gives Your AI Coder a Memory of Your Work

OpenAI shipped 'Chronicle' as an opt-in research preview inside the Codex Mac app on April 21. It continuously captures your screen, extracts text from what it sees, and stores that context locally — so when you start a Codex session, it already knows what you were working on, what errors appeared in your terminal, and which files you had open. The goal is to solve one of the biggest friction points in AI coding: re-explaining context every time you start a new session.

How Chronicle Actually Works

Chronicle runs a background process that takes regular screenshots and sends them to OpenAI's servers for text extraction. The extracted text — code snippets, error messages, file names, terminal output — is stored in a local database on your Mac. When you open Codex and ask a question, Chronicle injects relevant context from its memory automatically. You don't need to paste error messages or explain what you were doing. Codex already knows.

The storage is local and unencrypted. The screenshot processing happens on OpenAI's cloud infrastructure. This two-part architecture means your raw screen images leave your machine, but the extracted text stays on your computer.

What This Changes in Practice

The practical benefit shows up in these scenarios: you saw an error this morning and closed the terminal — Chronicle can recall the exact message. You had three documentation tabs open — Chronicle knows which APIs you were researching. A bug appeared while testing — Chronicle saw the state of your app before and after the crash. Each of these normally means copying and pasting context by hand.

I was debugging a failing test earlier today. Based on what you saw on my screen, what was the test checking and what error appeared? Can you suggest what the root cause might be?

What files have I had open most often this week? Based on the code you've seen, what area of the codebase am I primarily working on right now?

The Privacy Trade-offs

Screenshots processed on OpenAI's servers is a real concern for anyone working with proprietary business logic, customer data, unreleased product features, or code covered by an NDA. OpenAI's data processing agreement governs how screenshots are handled, but screenshots leave your machine — that's a straightforward fact. The local unencrypted storage is also worth noting: anyone with access to your Mac can read Chronicle's stored memories.

How It Compares to Microsoft Recall

Chronicle is OpenAI's answer to Microsoft's controversial Recall feature for Windows, which also captured screen content for AI memory. Recall faced significant backlash and multiple delays over privacy concerns. OpenAI made Chronicle opt-in with clear manual controls — a more cautious rollout — but the fundamental question is the same: do you trust cloud-processed screenshots in exchange for better AI context? The answer depends entirely on what's on your screen.