The AI agent (in Agent Mode or Composer) gets permanently stuck at the “Planning next moves” (or “Planning next actions”) stage, with the animation running forever, without proceeding, throwing an error, or completing the request. This happens on simple or complex prompts (e.g., “2+2” or code tasks like “generate a FastAPI endpoint”). No Request ID is generated because it never finishes – it just spins the loader indefinitely.
This started about 3 days ago and persists across all models tested . It’s not specific to one provider.
What I’ve Already Tried (without success):
Switched networks/internet (WiFi > 4G > Ethernet, DNS to 8.8.8.8).
Restarted the app, cleared cache (Settings > Advanced > Clear Cache), killed processes in task manager.
Reinstalled Cursor from scratch (uninstalled + fresh download).
Ran Network Diagnostics (Settings > Network): All OK (low ping, no connection/VPN errors).
Disabled Privacy Mode, switched models/providers multiple times.
Checked API/sub limits (Pro+ active, no rate limits hit).
No VPN, firewall temporarily disabled, OS updated.
This is killing my workflow – can’t code without the agent. Anyone else hitting this? Team, this seems like server overload or a bug in the planning logic? Please prioritize!
Steps to Reproduce
Open Cursor.
Go to Chat/Composer/Agent Mode.
Type any prompt (e.g., “Explain how an inclined plane works in Physics”).
Submit – the agent shows “Planning next moves” and animates, but hangs forever (tested for >10min).
Try canceling/restarting: It sometimes responds once, but hangs again on the next prompt.
Operating System
Windows 10/11
Current Cursor Version (Menu → About Cursor → Copy)
Hey, thanks for the report. This looks very similar to a recently solved case where extensions were causing the same infinite “Planning next moves” delay, even though network diagnostics suggested otherwise.
Try this first:
Run cursor --disable-extensions in CMD or terminal
In the new window that opens, don’t open a project - instead open Agent chat directly
Check if the agent responds
If it works, go to the Extensions panel and remove extensions one by one to find the culprit
It still doesn’t work and I don’t have any MCP servers configured and I’ve already tried to open different projects. Locally the AI Agents is working normally, the problem is when i connect to SSH the agents starts to slow down, planning next moves forever and when it works it comes to connection error.
WARN [ms-python.python]: Cannot register ‘python.venvPath’. This property is already registered. (workbench.desktop.main.js:55)
3. Warnings de Extensões (API Proposals Ausentes – Comum em Beta)
Via ‘product.json#extensionEnabledApiProposals’ extension ‘ms-toolsai.datawrangler’ wants API proposal ‘debugFocus’ but that proposal DOES NOT EXIST. (Repetido para: ms-vscode.vscode-copilot-data-analysis (‘chatVariableResolver’), ms-python.python (‘terminalShellType’), ms-python.vscode-python-envs (‘terminalShellType’), github.copilot-chat (‘chatReadonlyPromptReference’), ms-azuretools.vscode-azure-github-copilot (‘lmTools’), vscjava.vscode-java-pack (‘lmTools’), apidev.azure-api-center (‘chatParticipant’ e ‘languageModels’)) Stack comum:
WARN Failed to get current repository URL: Error: No full commit provider registered Stack similar, com h0o.getCurrentRepoUrl (8903:35478), getSlackInstallUrl (8903:35676), async u (2806:146577)
9. Erros de Composer (Direto no Planning)
[composer] Error in getPopulatedContext: EntryNotFound (FileSystemError): Error: ENOENT: no such file or directory, stat ‘/root/.cursor/projects/root/terminals/2.txt’ (workbench.desktop.main.js:4075, anonymous)
10. Outros Warnings/Erros Menores
Automatic fallback to software WebGL has been deprecated. Link: about:flags#enable-unsafe-swiftshader (workbench.html:1)
GL Driver Message (OpenGL, Performance, GL_CLOSE_PATH_NV, High): GPU stall due to ReadPixels (Repetido, workbench.html:1)
Thanks for the detailed console logs - they clearly show the issue! Your VPS can’t resolve api2.cursor.sh (DNS lookup failing with ENOTFOUND). This explains why it works locally but not on your remote server.
Could you run these commands on your VPS to confirm?
nslookup api2.cursor.sh
or
dig api2.cursor.sh
If DNS lookup fails:
Check if your VPS has firewall rules blocking certain domains
Try changing VPS DNS to 8.8.8.8 or 1.1.1.1 (Google/Cloudflare DNS)
Contact your VPS provider - some hosting providers restrict DNS resolution or have network policies blocking certain domains
If DNS lookup succeeds (domain resolves but Cursor still fails):
Share the output of the DNS commands above
Let me know your VPS provider and location
Troubleshooting the extension won’t help here since your local Cursor works fine - this is specifically a network connectivity issue from your VPS to Cursor’s servers.
Note: The only difference is i worked with a Network Share Folder that is opened into Cursor, but never get an issue and don’t think it’s related to this.
I’ve installed a previous version 2.0 and same issue but.. opening a working folder on a network shared folder and working from it seem’s to be the issue.
I’ve opened the same folder into my own PC which is on a local HDD into Laragon\www\ is working fine.
See the context chat, the header resume that the AI are generated the title of the chat resume, but not been able to working further (in Green the resume of the chat request) and (in Red the error stucked “Planning next moves”).
