Chronicle Setup Guide

Chronicle is in an opt-in research preview and is only available for ChatGPT Pro subscribers on macOS. It is not yet available in the EU, UK, or Switzerland.

Chronicle augments Codex memories with context from your screen so prompts can pick up work-in-progress context from recent activity.

Before enabling it, review the privacy and security section and understand the risks.

How Chronicle helps

Chronicle can reduce the amount of context you need to restate when working with Codex.

In practice it helps with:

• Using what is visible on screen so Codex can follow what you are currently looking at
• Filling missing context when you jump between tasks
• Remembering tools and workflows that you repeatedly use

In the right cases, Codex uses Chronicle context to identify the best source, such as a file, Slack thread, Google Doc, dashboard, or pull request.

Enable Chronicle

1. Open Settings in the Codex app.
2. Go to Personalization and enable Memories.
3. Turn on Chronicle below Memories.
4. Review the consent dialog and choose Continue.
5. Grant macOS Screen Recording and Accessibility permissions when prompted.
6. When setup completes, choose Try it out or start a new thread.

If macOS reports that a permission is denied, open System Settings > Privacy & Security and enable Codex under Screen Recording and Accessibility.

If a permission is restricted by macOS or your organization, Chronicle will start once the restriction is removed and permissions are granted.

Pause or disable Chronicle

You can pause and resume Chronicle from the Codex menu bar icon at any time.

• Use Pause Chronicle before meetings or when viewing sensitive content.
• Use Resume Chronicle when you want context to be captured again.

To disable it permanently:

1. Open Settings > Personalization > Memories.
2. Turn off Chronicle.

You can also control whether memories are used on a per-thread basis from the memories settings docs.

Rate limits

Chronicle runs background agents that summarize captured screen images into memories, and these agents can consume rate limits quickly.

Privacy and security

Chronicle uses screen captures and does not have access to microphone or system audio.

Where it stores data

• Temporary screen captures appear under $TMPDIR/chronicle/screen_recording/ while Chronicle is running. Frames older than 6 hours are deleted while running.
• Generated memories are saved in markdown files under $CODEX_HOME/memories_extensions/chronicle/ (usually ~/.codex/memories_extensions/chronicle).

Both locations can contain sensitive information. Do not share them with others.

You can ask Codex to search these memories. If you want Codex to forget something, delete or edit the respective markdown file.

What data gets shared with OpenAI

Chronicle captures are processed locally first, then summarized by Codex using selected screenshot frames, OCR text, timing, and local file paths.

Temporary screen captures are processed on OpenAI servers only for memory generation. OpenAI does not store screenshots after processing unless required by law, and does not use them for training.

The generated memories stay local in $CODEX_HOME/memories_extensions/chronicle/.

Relevant memory contents can be included as context in future sessions.

Prompt injection risk

Chronicle increases prompt injection risk from screen content. If you open a site with malicious instructions, Codex can be tricked into following them. Be cautious when running Chronicle in high-risk browsing environments.

Troubleshooting

I do not see the Chronicle setting

1. Confirm you are on a Codex app build that includes Chronicle.
2. Confirm Memories is enabled in Settings > Personalization.
3. Confirm Chronicle is available for your region and subscription tier.

Setup does not complete

1. Confirm Codex has Screen Recording and Accessibility permissions.
2. Quit and reopen the Codex app.
3. Open Settings > Personalization and check Chronicle status.

Which model is used for Chronicle memories

Chronicle uses the same model as your other memories.

If you did not set a specific model, it uses the default Codex model. To pin one, set consolidation_model in configuration.

[memories]
consolidation_model = "gpt-5.4-mini"