@Colin I’m hitting a closely related regression with the new Remote SSH Agent Host topology.
In my case the remote anysphere.cursor-agent-host successfully reaches Cursor, but Cursor returns:
ERROR_UNSUPPORTED_REGION This model provider is not supported in your region.
The same account/model works normally in a local Cursor workspace. The failure occurs specifically when Agent runs through Remote SSH.
I’ve already opened support ticket T-F93858, but the AI support assistant treated it as a standard regional-model-availability issue and only suggested using Auto. I’m not requesting a regional exception — I’m trying to get the Remote SSH / Agent Host regression investigated.
Example Request ID: 4de2316a-7783-4af1-83fd-892c3dbcad5e
Could you please help route this to the Agent Host / Remote SSH team, or let me know where I can privately send my Cursor account email and the remaining Request IDs for debugging?
I’m experiencing the same issue with Cursor over Remote SSH.
Everything worked normally until about a week ago. My local Cursor works fine with Claude/GPT, but when I connect to my remote Linux server via SSH, Agent fails with:
[permission_denied] Cursor is not available in your region
The remote server is located in Russia, while I’m using Cursor locally from a supported region.
This only started recently, around the time of the new Remote SSH / Agent Host changes. Is it possible to keep AI/model requests running locally while using the remote server only for files, terminal and execution, like before?
This workflow is critical for me because I cannot move the server.