Sync ChatGPT Project Memory with Codex for Seamless Ideation-to-Code Continuity

Resolved 💬 12 comments Opened Oct 13, 2025 by thetopham Closed Nov 16, 2025
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What feature would you like to see?

🧠 Feature Request: Sync ChatGPT Project Memory with Codex

Summary
A lot of idea creation, planning, and architecture design happens inside the ChatGPT chat interface — not just in the code editor. These insights are stored in ChatGPT’s project memory, but Codex (or the ChatGPT code editor) currently has no access to that memory.

Problem
When switching from ChatGPT conversations to coding in Codex, all the reasoning, naming conventions, and design context discussed earlier are lost. This breaks continuity between brainstorming and implementation.

Requested Feature
Add a way for Codex to access and sync ChatGPT project memory, so it can recall:

  • Design and architecture decisions discussed in chat
  • Feature ideas, goals, and iteration history
  • Naming conventions, TODOs, and style preferences

Possible Implementation Ideas

  • A “Sync Chat Memory” button in Codex that imports relevant chat/project memory
  • Automatic linking when a repo is associated with a ChatGPT project
  • A sidebar or tooltip summary showing context from past chats

Benefits

  • Maintains creative continuity between ideation and coding
  • Makes Codex a true extension of ChatGPT rather than a separate tool
  • Reduces repetitive explanations and improves long-term project reasoning

Additional information

_No response_

View original on GitHub ↗

12 Comments

github-actions[bot] contributor · 9 months ago

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

  • #4268
  • #4933
  • #4655
  • #4290

Powered by Codex Action

interconnectedMe · 8 months ago

Would like this

rmasciarella · 8 months ago

it must already do this. I will code something in codex and then the next day have a pulse that is an enhancement on that exact feature with my projects exact context.

jonahdk · 8 months ago

@rmasciarella I don't think that's what they're talking about. First, Pulse is only available for Pro. Second, they never mentioned Pulse, just Projects. I think what they mean is so that you can just use ChatGPT in Projects for higher-level plans instead of having to use Codex, so that Codex can stick to lower-level plans closer to code (since ChatGPT is generally better at higher-level planning in my experience, and to preserve Codex usage limits)

interconnectedMe · 8 months ago

I am not 100% sure of the OP's requirements.

As for myself, this is not related to limits. Maybe for the OP, but it doesn't sound like it to me.

In my usage, this is simply about 'ChatGPT in e.g. VScode > Codex' knowing about other chats it has with me....in VScode/web/phone app/VScode > Codex.

Even when working in VScode with Codex, then starting a new chat, there is not continuity between chats. I will ask in Codex chat if it is aware of the other chat containing XYZ steps to continue, and it states it is not aware.

So we have to go around in circles a bit with ChatGPT's own suggestions of copying and pasting key points, or creating .md files, pinning them and telling me, to tell it, in the new chat, to reference that pin, then not remembering that it (ChatGPT) itself (in VScode > Codex Extension) had already suggested and setup scripts to export chats and save in /docs as .md files to refer back to.....and other documentation as 'source of truth'..... and it all ends up being just a bit messy.

If this is non-extant / or / a bit buggy functionality, it would probably be far simpler for ChatGPT in Codex, just to be able to see 'Projects' which contain specific Chats/(Tasks in VScode Codex sidebar for me). These could simply match the 'Projects' I have in ChatGPT web. Then it has a concrete set of data to reference (rather than it having to do something 'smart' like infer what chats (Tasks in Vscode Codex sidebar) are relevant.

Any ideation chats done via web, or app, or wherever, could be moved by me into the Project it relates to (whatever platform I am using at that time), and viola....tada, it's in the concrete set of chat / tasks under that project when I start VScode and look in Codex sidebar tasks.

I think that makes sense to me anyway.

etraut-openai contributor · 8 months ago

Closing in favor of #2513

lefty3382 · 8 months ago

this issue is exactly what I was searching for. I created a project in ChatGPT with multiple conversations regarding strategy around network deployment, and then I wanted to share that chat history with a Codex environment to give context to the tasks I now want it to perform, but that isn't possible. So now I have to be the gopher between my ChatGPT project discussion history and the Codex task. There should be a way to link the two together.

interconnectedMe · 8 months ago

I don't understand why was closed 'in favour of 2513'? It certainly does not seem to be a duplicate, either.

This issue's feature request title - "Sync ChatGPT Project Memory with Codex for Seamless Ideation-to-Code Continuity #5152" -
is not related to the usage limits in #2513 (better information about usage limits)

This issue seems to be that ~

OBSERVED BEHAVIOUR:

  1. Plan a Project and ideate in ChatGPT (browser)
  2. Go to VScode > Codex Extension sidebar to begin work on Project that was planned in ChatGPT (browser)
  3. Project is not present in Codex Extension.
  4. Codex does not see that Project, so start manually copy & pasting from browser....

EXPECTED / DESIRED (?) BEHAVIOUR:

  1. Plan a Project and ideate in ChatGPT (browser)
  2. Open VScode, with Codex Extension installed.
  3. Continue Project implementation, being able to view the Project Chats from ChatGPT (browser).
  4. This does not exist, but would, I think, when I click on Task History (in the Codex Extension Sidebar), drop down as a natural child-menu item of 'Local'.
  5. Codex Extension can then be referred to relevant chats for context / decisions, or at least manual copy-paste is a little less clunky.
etraut-openai contributor · 8 months ago

Apologies, I transposed the number in the other issue. I should have typed #2153

Hung-Nguyen-tfs · 8 months ago

Why shut down the discussion when people are trying to solve the problem? Consolidation can come later — killing momentum now just hurts progress.

etraut-openai contributor · 8 months ago

@Hung-Nguyen-tfs, we generally use upvotes to prioritize features. Duplicate (or largely overlapping) features dilutes the signal for a given feature. If you want to see a particular feature idea prioritized, consolidation will help.

Hung-Nguyen-tfs · 8 months ago

I get the intent behind consolidation, but active discussion threads often surface valuable insights and edge cases that get lost when merged or closed. Upvotes alone don’t capture that nuance — sometimes momentum matters more than a tidy feature list.