Advanced Project Memory Management

Feature Request: Advanced Memory Management for Projects

Hello OpenAI Team,

I’d like to suggest an improvement to the ChatGPT Projects feature.

Currently, Projects provide a useful way to organize conversations around specific topics. However, I believe the memory system could be much more flexible, allowing users to control how each Project interacts with their global ChatGPT memory.

My suggestion is to introduce three configurable memory modes for each Project:

1. Shared Memory (Read & Write)

  • The Project can access information from the user’s global memory.

  • Information learned or established within the Project can also contribute to global memory.

  • This provides seamless interaction between Projects and regular conversations.

2. Shared Memory (Read-Only)

  • The Project can access and use the user’s global memory.

  • However, information, preferences, and context established within the Project must remain within that Project.

  • Nothing from the Project should automatically be added to global memory.

This is particularly useful when working on specialized topics without mixing their context with the user’s general profile.

3. Isolated Memory

  • The Project has its own independent memory.

  • It cannot access global memory.

  • Information from the Project cannot be added to or influence global memory.

  • Conversations within the same Project can still share context with one another.

An Additional Suggestion: Selective Memory Sharing

It would also be useful to allow users to manually choose which information from a Project can be transferred to global memory, rather than automatically sharing everything.

Why Would This Be Useful?

For example, a user might create separate Projects for:

  • JavaScript and Web Development

  • Learning English

  • Linux and Networking

  • Personal Research

  • Professional or academic work

Each of these areas may require different instructions, context, preferences, and accumulated knowledge.

The current Project-only memory option is useful, but it doesn’t fully address the need for a middle ground: using global memory without allowing project-specific information to flow back into it.

I believe this would make Projects significantly more powerful and flexible, especially for users who use ChatGPT as a long-term learning, research, and productivity environment.

Thank you for considering this suggestion. I hope this feature can be considered for a future release.