Hey, this is a known intermittent issue that multiple users reported around the same time. It’s likely related to temporary capacity pressure on April 16.
A couple of things to try:
Run network diagnostics: Cursor Settings > Network > Run Diagnostics
Try disabling HTTP/2: App Settings Cmd+, > search HTTP/2 > enable Disable HTTP/2. This often helps on certain networks.
Try selecting a specific model instead of Auto to see if the issue still happens with particular models
Also, can you share a Request ID from one of the failed requests? You can get it from the chat context menu top right > Copy Request ID. This will help us check what’s happening on our end for your account.
Hi @deanrie thanks for the reply, …here is the result:
network diagnostic, all OK including HTTP/2, but disabled anyway
selecting a specific model instead of auto: it helped somewhat (not Claude model though, but GPT-5.4), at least the task did not fail immediately, and started to execute, did some code updates too, but stuck again and then “Agent is waiting for a command to finish” is running forever.
Here is an ID: `Request ID: 01935a4a-065e-4385-b1c8-70bf4e32c903`
About “Agent is waiting for a command to finish”, that’s a separate issue from connection failed. Can you confirm what command the agent was trying to run at that moment? If a terminal command got stuck, try clicking Cancel or opening a new chat.
If the connection failed issue is still happening today, let me know. On April 16 there were lots of reports about connectivity issues to providers, so it’s likely things have stabilized now.
If after disabling HTTP/2 the diagnostics show Failed to fetch for Downloads and CDN, try turning HTTP/2 back on. It looks like HTTP/2 works fine on your network, and disabling it is what breaks the connection.
Provider Error for models is most likely a temporary provider-side issue, there were a lot of reports on April 16. Can you share your Cursor version and OS? Also, does it still reproduce today?
I was under the impression that I have sent another reply, …never mind. …you were right, it seems I had two issues at the same time, and one shaded the other. I had some iPV6 issue, and my network settings were corrupted. I haven’t noticed until several of my apps started to fail to connect, like Signal, Chrome, beside Cursor. …yes, all Chromium, and Safari was up and running. …anyhow, after some cleanup Chrome is back, but most importantly for this forum, Cursor is also back and the agent is working as expected