[Feature request] Let API apps use a user's ChatGPT account context (Memory, Projects, connected apps)

TL;DR

Let users OAuth-authorize API apps to access the personal context already in their ChatGPT account — Memory, selected Projects, relevant chat history, and selected connected apps — with optional write-back. Users shouldn't have to rebuild and synchronize a second personal database just to use their own accumulated context in wearables, agents, local models, custom apps, or alternate I/O such as smart glasses.

The missing primitive is essentially:

openai.responses.create(
    use_my_chatgpt_memory=True,
    use_my_connected_apps=True,
    use_my_projects=True,
)

Summary

Please add an opt-in, user-authorized API for ChatGPT account context so an external app can use the same personal context the user has already built inside ChatGPT: Memory, selected Projects, relevant chat history, and selected connected apps.

The missing primitive is roughly:

openai.responses.create(
    model="...",
    input="...",
    use_my_chatgpt_memory=True,
    use_my_connected_apps=True,
    use_my_projects=True,
)

Not necessarily with that exact API shape — OAuth-style scopes would probably be better.

Problem

ChatGPT is increasingly becoming a user's personal context layer:

• long-term Memory
• years of conversation/project history
• ChatGPT Projects and their files/instructions
• Gmail / Drive and other connected apps
• newer personal-data integrations such as Health and Finances

But that accumulated context is largely usable only inside ChatGPT.

If I build another interface — for example a wearable assistant, custom mobile app, local agent, voice interface, or ambient-computing system — the OpenAI API does not appear to have a documented way for the user to authorize that app to use their existing ChatGPT account context.

That forces users to either:

1. rebuild a second personal database;
2. separately reconnect every upstream service;
3. continually synchronize the two systems; or
4. give up the personalization they already accumulated in ChatGPT.

The result is that the user's data/profile is fragmented by application even when the user wants the same AI context across surfaces.

Requested capability

Allow a user to OAuth-authorize an API application to retrieve relevant ChatGPT account context for a request.

Possible scopes:

chatgpt.memory.read
chatgpt.memory.write
chatgpt.history.retrieve
chatgpt.projects.read
chatgpt.projects.files.read
chatgpt.connected_apps.use

Sensitive connected sources could retain their own narrower scopes and restrictions rather than being implicitly exposed.

Conceptually:

                     User's ChatGPT context
                    /        |         \
              Memory      Projects    Connected apps
                    \        |         /
                     authorized context API
                              |
              +---------------+---------------+
              |               |               |
          wearable app     local agent     custom UI
              |               |               |
              +-------- OpenAI Responses API--+

The API should retrieve only context relevant to the current request, rather than dumping the user's entire account history into the model context.

Example use case

I am building a personal wearable/ambient assistant.

The glasses/phone handle realtime microphone/camera input and call a frontier model. I want the interaction to be able to answer things like:

• "What was the name of that thing we discussed last week?"
• "What did I decide about this project?"
• "Do I have an email related to this?"
• "Remember this for later."

The useful context already exists in ChatGPT, but today the external application would need to recreate the user's identity/memory layer from scratch.

Instead, the architecture should be:

wearable / app
      |
OpenAI API + user-authorized ChatGPT context
      |
Memory / Projects / selected connected apps

Then the physical interface and even the model can change without forcing the user to rebuild their personal context.

External I/O is the missing piece

Another way to frame this request: ChatGPT already has the user account and context I want; I just want to attach additional I/O to it.

On a phone I can sign into ChatGPT and interact through the official app, including voice/vision-style interfaces. But if I want smart glasses, a wearable microphone/camera, a custom HUD, hardware buttons, or another client to interact with that same accumulated context, there is no equivalent user-account API surface I can authorize.

I do not necessarily want to replace ChatGPT or maintain a second memory system. I want something closer to:

smart glasses / wearable I/O
          |
     phone companion
          |
 user-authorized ChatGPT account API
          |
Memory / history / Projects / connected apps

The programmable API and the ChatGPT consumer account currently feel like separate context planes. A bridge between them would let developers build new interfaces around the user's existing ChatGPT context instead of exporting, recreating, and continuously synchronizing it elsewhere.

Ideally the same authorization could work with Responses and realtime voice/multimodal sessions, so the interface can change without losing the user's accumulated context.

Write-back would make this much more powerful

Ideally this is bidirectional.

A conversation that occurs through an authorized wearable/app could be eligible for explicit memory write-back, so information learned outside the ChatGPT UI can later be available inside ChatGPT.

For example:

user_context={
    "memory": "read_write",
    "projects": ["Manfred"],
}

with clear user controls over what is persisted.

Privacy / control requirements

This should be user-owned and permissioned rather than giving every API key automatic access to a ChatGPT account.

Suggested properties:

• explicit OAuth consent
• least-privilege scopes
• per-Project allowlists
• separate handling for sensitive sources
• app-level revocation
• visible audit/history of context access
• optional read-only vs read/write Memory access
• no access by default
• retrieved-context limits / relevance filtering

Why now

This request is broader than simply "add Memory to the API." As ChatGPT accumulates Projects, connected apps, and user-specific context, the value is increasingly in the combined personal context layer.

Without an authorization bridge, users who build their own interfaces are pushed toward duplicating that entire layer elsewhere.

I would rather authorize my applications to use my existing context than have every company and every assistant maintain a separate profile of me.

Related discussions include the older openai/openai-python#1699 Memory API request and current Codex/ChatGPT shared-memory requests such as #27121 and #30159, but this proposal specifically asks for a general account-context authorization surface for API applications.

I realize openai/codex may not be the ideal tracker for a cross-ChatGPT/API product request; I could not find a dedicated public ChatGPT feature tracker. Please route it internally if there is a better destination.

cc @thsottiaux — this seems particularly relevant as ChatGPT and Codex converge.