VS Code extension: add a setting to disable cloud conversation fetching
What feature would you like to see?
Please add a documented VS Code/Codex setting that disables background fetching and synchronization of cloud conversations/tasks.
For example:
{
"chatgpt.cloudConversations.enabled": false
}The exact name is not important, but the setting should provide a supported local-only mode for conversation history.
When disabled, the extension should:
- Stop listing, refreshing, subscribing to, and retrying cloud conversation/task-history requests.
- Continue to show and support local Codex sessions.
- Continue to allow the network calls required for normal model requests and authentication.
- Explain in the UI that cloud history is disabled by the user, rather than showing repeated fetch errors.
- Optionally allow a one-time manual refresh or an explicit switch back to cloud history.
Current behavior
<img width="1195" height="122" alt="Image" src="https://github.com/user-attachments/assets/11f8c873-9ea9-4359-b356-f032ad2b4877" />
In a restricted network environment, a firewall blocks the cloud conversation/history endpoints. The Codex VS Code extension continues attempting to fetch them and repeatedly reports failures.
There does not appear to be a documented setting that disables only cloud conversation fetching. Blocking broader OpenAI/ChatGPT endpoints is not a practical workaround because those endpoints may also be required for authentication and normal Codex requests.
Why is this needed?
This would help users working:
- Behind corporate or allow-list firewalls.
- In restricted or intermittently connected development environments.
- With a local-session-only workflow.
- In environments where cloud conversation synchronization is intentionally disabled for privacy or policy reasons.
It would also avoid repeated failed requests, noisy logs/errors, and unnecessary retries when the user already knows cloud history is unavailable.
Suggested behavior
A workspace- or user-level setting could disable only the cloud conversation/history feature. If the setting is off, the extension should not start the related fetch or subscription loops.
As a secondary improvement, failures should use bounded exponential backoff and degrade gracefully instead of retrying continuously.
Environment
- VS Code:
1.130.0 - Codex extension:
openai.chatgpt@26.5721.30844 - Platform: Windows x64, build
26200
Additional information
I am happy to provide sanitized logs or test a proposed setting. The main request is for a supported, narrowly scoped switch rather than requiring users to block shared endpoints at the firewall.