I use ChatGPT to maintain an active community project in Google Drive. Some committee members do not use Google accounts, so I would like ChatGPT to maintain selected documents in OneDrive or SharePoint as a read-only committee mirror.
The missing capability is not just reading OneDrive. ChatGPT needs to be able to create, upload, replace and update files there on an ongoing basis.
Ideally, I could approve:
a source folder in Google Drive,
a destination folder in OneDrive/SharePoint,
which files are included,
and a sync rule.
Sharing permissions should remain under the user’s control, with the destination able to stay view-only for committee members.
This would be very useful for committees, boards, clients and mixed Google/Microsoft organizations.
I’d like to request support for event-triggered ChatGPT Scheduled Tasks or automations, where an external authorized system could trigger a task immediately through a webhook or similar event mechanism.
My use case is a project workflow where an external process can detect a new event within about one minute, but ChatGPT’s scheduled processing currently has to wait for its next time-based run.
A webhook/event-trigger option would allow the external system to signal that work is available, while ChatGPT would still perform the actual authorized processing using the task’s existing instructions and connected tools.
Ideally, this would support:
- A secure webhook or event endpoint tied to a specific Scheduled Task.
- Authentication/authorization so arbitrary external callers cannot trigger tasks.
- Event payloads or a small amount of trigger context.
- Existing task instructions, permissions and connected-tool boundaries remaining in force.
- Deduplication/idempotency support so repeated webhook delivery does not cause duplicate processing.
- Logging of trigger time, source and execution result.
- Rate limits and abuse protection.
This would be useful for workflows involving incoming email, form submissions, CRM changes, project-management events, file changes and other situations where an external system can detect the event immediately but ChatGPT is responsible for the higher-level reasoning and follow-up.
It would also allow users to keep lightweight external detection systems while avoiding frequent polling simply to determine whether there is work to do.
A general event-trigger capability for Scheduled Tasks would make ChatGPT much more useful as an operational automation layer while still preserving user-controlled permissions and task boundaries.
I’d like to request a connector for the OpenAI Developer Community, or more generally for Discourse-based forums.
A connector like this would allow ChatGPT to work directly with authorized community forum content instead of requiring users to manually copy threads, replies, and URLs between the forum and ChatGPT.
Useful capabilities could include:
- Search existing forum topics before creating a new post.
- Read topics, replies, moderator responses, and status changes.
- Create a new topic after explicit user approval.
- Reply to an existing topic after explicit user approval.
- Follow or monitor selected topics for meaningful new responses.
- Retrieve the canonical thread URL and topic metadata.
- Preserve the existing forum thread when posting follow-up information.
- Respect the user’s existing Discourse/OpenAI Community account permissions.
- Clearly separate read actions from write actions and require appropriate authorization for posting or editing.
One practical use case is feature-request management. A user could ask ChatGPT to research whether a similar request already exists, prepare a non-duplicate request, post the approved version, and later monitor that thread for replies or status changes.
More broadly, a Discourse connector could be useful across many technical communities that use the platform, rather than supporting only one forum.
The connector should ideally use authenticated APIs rather than browser automation and maintain normal ChatGPT permission, privacy, and confirmation boundaries.
I ran into a closely related problem with OneDrive Personal (MSA) and ended up finding a reproducible workaround using the exact Graph-backed primitives currently exposed by ChatGPT’s Microsoft connector.
On the tested personal account, some convenient high-level actions such as native folder creation, search/discovery and delete were unavailable, but exact copy, move/rename and file-content replacement operations worked.
The workaround uses a one-byte file seed and a folder seed to reconstruct create-folder/create-file semantics through copy → Graph-path resolution → exact overwrite.
I have now reproduced it twice from scratch and used it for two real document-store migrations. Across those migrations, 13/13 pre-existing TXT/PDF/JPEG/ZIP files were downloaded from the source, SHA-256 hashed, written to OneDrive, downloaded again and verified byte-for-byte.
I documented the procedure, limitations and validation in the public GitHub repository:
mattek8/chatgpt-onedrive-personal-workaround
This is not an OAuth bypass or security vulnerability — it only composes operations already authorized by the connector. It seems more like an MSA compatibility gap between the higher-level connector actions and underlying Graph capabilities.
Hopefully this is useful both as a temporary workaround and as evidence that a cleaner native writable OneDrive Personal connector path should be possible.
Thanks for these detailed ideas! SharePoint can now create/update files when the needed actions are enabled: https://help.openai.com/en/articles/12143177. Some app-event tasks also exist, but the automatic cross-cloud mirror, general webhook and Discourse workflows you describe aren’t fully covered by the docs. We can pass those requests along; no timeline to share.