Hi all,
Starting a few days ago, I’ve noticed that my Custom GPT Actions consistently fail without ever reaching my server. These same Actions used to work fine for months.
What’s happening:
-
The GPT hangs for over a minute when the Action is triggered.
-
Eventually, it fails with the message:
An error occurred while executing the plugin.
-
Most importantly, no request ever reaches my server — no logs, no DNS lookup, nothing.
-
I’ve tried:
- Multiple different domains (with and without HTTPS)
- Two completely different servers
- Super simple GET endpoints (
?hashtag=travel
) that work fine outside of GPT (tested in browser and curl)
What I’ve ruled out:
- This only happens within Custom GPTs — the endpoints are fully reachable otherwise.
- I double-checked server access logs — no trace of incoming requests from OpenAI’s infrastructure.
- I suspect it might be a DNS issue or internal timeout before the request even goes out.
What I’m wondering:
- Has OpenAI introduced any new restrictions on outbound Action calls?
(There is a new “Recommended Model” section now…) - Are certain domains, IPs, or DNS providers being blocked or filtered?
- Is there any way to debug these failures more deeply from the GPT side?
I’ve attached a screenshot of the error for context.
Any help or confirmation would be appreciated — happy to DM my domain/server details if needed.