Hi, I’ve been using Cursor with a specific inference provider, and I noticed that when the provider returns a 429 error in the middle of an agent workflow, Cursor simply shows an error and I have to restart the entire process.
Is there any option to resume the workflow after this kind of error?
I also created a proxy server to check how many retries were being made, and I saw only one retry for the 429 error. Is that expected behavior?
Hi Jiyoon,
Yes, seeing one automatic retry is expected for this custom-provider path. After that retry, Cursor surfaces the 429 rather than continuing indefinitely. The retry count and backoff are not currently configurable.
You do not need to start a new conversation. Once the provider is available again, you can continue in the same Agent thread. Use Try again or Resume if shown; otherwise, send “continue” or resubmit the last instruction. Earlier conversation context and file edits remain available.
Automatic resume after the retry is exhausted is not currently supported, and custom-provider requests do not fail over to another provider.