My client has been using a custom connector for quite some time, ever since the custom connector beta. It’s now an internal app used in their business workspace. Recently, the AI client has begun asking for the users for permission to run the tool calls, even though it never did this before. That is too disruptive to their workflow and prevents useful use of ChatGPT.
The tools are annotated with read-only, etc., so there should be no problem here and it used to run just fine. Why did this start happening, and are there any settings or additional configurations that we can do to prevent it? I’ve tried the other annotations as well, but nothing fully prevents it, and it seems to happen haphazardly. This is a custom internal app in a business workspace.