Today the service quality has degraded significantly. When you create a worktree or work in the cloud and then bring it locally, the session connection is lost and cannot reconnect.
**OS:** Windows 11 (10.0.26200)
**Date:** 2026-09-15
**Expected:** The agent session stays attached (or reconnects) after the worktree is created or the cloud run is opened locally.
**Actual:** The session connection drops and reconnect fails.
**Related: Agent Execution Timed Out**
Dialog text: “The agent execution provider did not respond in time. This may indicate the extension host is not running or is unresponsive.”
Request ID: 16f46181-3648-4322-947f-b017f4ed1da1
**Related: Cannot start a chat in a local repo folder**
User report: “I can’t even start a chat in my local repo folder.”
Error: Agent Execution Timed Out — “The agent execution provider did not respond in time. This may indicate the extension host is not running or is unresponsive.”
Request ID: 2909aa2e-a51e-4e06-8b07-9a51db703e2f
Also previously seen Request ID: 16f46181-3648-4322-947f-b017f4ed1da1
Hi @Vince_Roche, thanks for the detailed report, and sorry for the disruption.
I checked the affected requests. This appears to be a local Agent startup issue when certain worktrees or cloud sessions are opened locally. It isn’t caused by anything in your setup.
I’ve made a change on our side that should prevent this after restarting Cursor. Please:
Open Task Manager and end every Cursor process.
Reopen Cursor.
Test a new local Agent chat.
Reopen one of the affected worktrees or cloud sessions locally.
Please let me know whether both cases now work normally.
If the issue continues, press Ctrl+Shift+P, run Capture and Send Debugging Data, and accept the defaults. Please also share the new Request ID, approximate test time, and affected project or worktree name. If the command saves a zip file to your Desktop, attach it here.
Quick follow-up on the “Agent Execution Timed Out” error when opening new worktrees or cloud sessions locally. We found the cause of this timeout in the newer agent runtime, and a fix should have been included in Cursor 3.21.18.
If you still see the timeout after a full restart, or once you’re on 3.21.18+, please share a new Request ID here and we’ll take another look.