Codex feedback: ChatGPT Project copying is triggering blocks

[EARTHTOEDWARD]

Summary

I am trying to copy my own ChatGPT Project conversations into my own local repository, /Users/edward/Desktop/Chatgpt_Library, as a personal archive. The task is internal data portability: copying chat titles/transcripts from my ChatGPT account into a local repo that I control.

The current Codex/browser workflow is now causing red-flag behavior in ChatGPT. Multiple conversation loads are returning Unable to load conversation, and prior fallback attempts opened unwanted additional ChatGPT instances in Chrome. The automated workflow should pause rather than keep retrying.

User Goal

• Preserve my own ChatGPT Project chats in a local repo.
• Keep the original ChatGPT project/chat folder structure.
• Copy only transcript text for now.
• Count a chat as copied only when content.md exists and is non-empty.

What Went Wrong

• Automated conversation-opening attempts began producing Unable to load conversation.
• Some direct conversation URL opens bounced back to the project page instead of rendering the transcript.
• Earlier fallback behavior opened extra ChatGPT instances in Chrome, even though the intended browser surface was the Codex in-app browser / ChatGPT Atlas path.
• The copy workflow became indistinguishable from repeated scraping/retry behavior, even though the user intent is legitimate personal data export.
• Continued retries risk triggering more security warnings and making the account/session less usable.

Why This Matters

This is not an attempt to copy third-party data or bypass access controls. The user is signed into their own ChatGPT account and is trying to preserve their own conversations. Codex needs a safer, first-class path for personal ChatGPT data portability that does not rely on brittle browser scraping or repeated conversation reloads.

Requested Product Fix

Please provide a supported Codex/ChatGPT workflow for exporting or locally archiving a user's own ChatGPT Project chats, including:

• A stable way to enumerate projects and conversations.
• A stable way to export transcript text for conversations the user owns.
• Clear rate limits or safety guidance before red flags are triggered.
• A way for Codex to distinguish legitimate personal archive/export work from suspicious automated scraping.
• Browser automation controls that do not silently switch surfaces or open unwanted Chrome/ChatGPT instances when the user requested the in-app browser or Atlas path.

Current Workaround Status

The repo-local copy should remain paused until the block clears or a safer export path is available. The latest progress note is:

/Users/edward/Desktop/Chatgpt_Library/Workflow/chatgpt_project_copy_progress_2026-05-06.md

At the pause point, the workflow had copied many non-pulse missing chats successfully, but the remaining ChatGPT Project copy attempts were increasingly causing Unable to load conversation blocks.

[github-actions]

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

• ChatGPT Data Controls export loops through MFA without creating export request #20999
• Feature Request: Create Codex Task from ChatGPT Conversation or Project #19788
• Feature Request: LLM Wiki / Auto-Index Layer for ChatGPT History and Codex Workspaces #20920
• Computer Use fallback after Browser Use failure should preserve requested browser/session preference #21068

Powered by Codex Action