OpenAI introduced Chronicle as a feature that helps Codex create memories using what’s visible on your screen. It reduces the need to manually provide context by enabling Codex to understand ongoing tasks, tools, and workflows with minimal input.
Chronicle and How Chronicle Works
Chronicle is designed to simplify how you interact with Codex by automatically building context from recent screen activity. Here’s how it works:
- Uses on-screen context to understand what you are currently viewing
- Fills missing context without requiring detailed prompts
- Learns tools and workflows over time
- Helps identify the correct files, documents, or sources
- Switches to better sources such as files, Slack threads, Google Docs, dashboards, or pull requests when needed
- Reduces repetition across prompts
Setup, Controls and Rate Limits
Enable Chronicle
- Open Settings in the Codex app
- Go to Personalization
- Turn on Memories
- Enable Chronicle
- Review the consent prompt and select Continue
- Grant Screen Recording and Accessibility permissions
- Choose Try it out or start a new thread
If permissions are denied, enable them via System Settings > Privacy & Security. Chronicle will start once permissions are granted.
Pause or Disable
Use the menu bar icon to pause or resume Chronicle. Pause it during meetings or while viewing sensitive content.
- To disable, go to Settings > Personalization > Memories and turn it off.
You can also control whether memories are used within a specific thread.
Rate Limits
Chronicle uses sandboxed background agents to process screen activity and generate memories. These agents can consume rate limits quickly.
Privacy and Security
Data Storage
Screen captures are stored temporarily on the device and deleted after 6 hours. Temporary files may appear under $TMPDIR/chronicle/screen_recording/.
Generated memories are stored locally as unencrypted markdown files in $CODEX_HOME/memories_extensions/chronicle/. You can view, edit, or delete them anytime, but should not manually add new information.
Data Usage
Chronicle captures screen context locally and creates memories using a temporary Codex session. It may process selected screenshots, OCR-extracted text, timing data, and local file paths.
Screenshots are processed on servers to generate memories but are not stored after processing unless required by law and are not used for training.
Memory content may be included in future sessions and can be used to improve models if allowed in your settings.
Prompt Injection Risk
Since Chronicle processes screen content, it increases exposure to malicious instructions from websites or applications.
Troubleshooting and Model Configuration
Ensure the Codex app version supports Chronicle and that Memories is enabled. Check Screen Recording and Accessibility permissions, and restart the app if needed. If permissions are restricted, Chronicle starts after access is granted.
Chronicle uses the default Codex model for memory generation. You can set a specific model manually:
- [memories]
consolidation_model = “gpt-5.4-mini”
Availability
Chronicle is available as an opt-in research preview in the Codex app on macOS for ChatGPT Pro users. It is not available in the EU, UK, or Switzerland.
It requires Screen Recording and Accessibility permissions. Before enabling, note that Chronicle can consume rate limits quickly, increases prompt injection risk, and stores memories locally in unencrypted form.