Hey, thanks for the report. From the logs for this request, it looks like the session ran for about 10 minutes and then failed due to a server-side timeout. That’s a typical sign of an unstable or dropped connection to our endpoints, not a model failure. The “it cuts out and I have to retry” symptom is known right now, and I can’t share an ETA for a general fix.
What to try:
Cursor Settings > Network > Run Diagnostics. This will show where the connection drops.
App Settings Cmd+Shift+, > search for HTTP/2 > turn on Disable HTTP/2. This often helps on unstable networks, VPNs, and proxies.
If you can, test on a more stable connection. The difference is usually noticeable.
One more thing. On the Free plan, if you go over your included request volume, requests can get delayed or queued, which can also look like disconnects. If that’s your case, switching networks won’t fully remove the delays.
Let me know what Network Diagnostics shows and whether disabling HTTP/2 helped.
I have configured and checked all the above steps. I have tried switching networks and turning off VPN, but it still doesn’t work. This situation has been occurring for the past two days and often prompts for reconnection.
Thanks for the screenshots, they help a lot. It looks like the diagnostic is failing on the Agent test [aborted], and Ping is showing very high latency 2066ms, while DNS, SSL, API, and Chat pass. That usually means it’s not a model issue, it’s a connection issue to the specific endpoint the agent uses. The agent relies on a persistent streaming connection, and that’s the first thing to break on an unstable or high latency network.
Since you already switched networks, turned off VPN, and disabled HTTP/2, and the other checks are green, this points to regionally high latency to our endpoint. We’re aware of this type of issue, but I can’t share a clear ETA for an infrastructure-side improvement yet.
What can actually help on your side:
Try again at different times of day. With regional latency, the difference can be pretty big.
If you can, test with another ISP or a mobile hotspot. That helps confirm if it’s tied to a specific route.
For long tasks, split the request into smaller steps. That lowers the chance the streaming connection drops during one long run.
Also keep in mind: on the Free plan, if you go over the included usage, requests can get delayed or queued, and that can look like disconnects. Switching networks won’t fix that.
If you find a network where the Agent check passes in diagnostics, let me know. That’ll help us narrow down the cause.
