[FEATURE] VS Code Extension: Local API proxy for Remote-SSH like Copilot
Preflight Checklist
- [x] I have searched existing requests and this feature hasn't been requested yet
- [x] This is a single feature request (not multiple features)
Problem Statement
As an ML researcher using a remote cluster, I am having problems with cluster firewall blocks api.anthropic.com:443 (ECONNRESET)—Claude Code fails. GitHub Copilot works perfectly: runs locally, agents edit/execute remotely.
This is critical due to the great computational requirements that it needs that are only supplied in the cluster.
<img width="428" height="167" alt="Image" src="https://github.com/user-attachments/assets/41161f22-5e7f-49d7-b2b3-527029957d7d" />
<img width="934" height="418" alt="Image" src="https://github.com/user-attachments/assets/200d4a93-4aab-4cec-9b15-103d5b06a180" />
Proposed Solution
I'd like to run VS Code extension proxy API calls locally (Copilot-style), while executing the changes in remote.
Alternative Solutions
As alternatives I've tried
- sshfs + local Claude: No live exec (local CPU explodes on ML training)
- Tinyproxy/SSH tunnel: Flaky (400/CONNECT fails)
- Admin whitelist: Slow/policy issues
Currently what has worked for me has been sshfs + local Claude, but I can't test the code since it heavily rely on computational and GPU resources that I do not have in my machine.
Priority
High - Significant impact on productivity
Feature Category
Other
Use Case Example
Example scenario:
- Install Claude Code CLI/extension locally + cluster (Remote-SSH).
- Try Claude Code in the cluster (If it works with no blocking, the personal workflow continues from here).
- If it fails in the cluster, I modify the JSON config to allow Claude Code to run on the local machine but having access to the remote files.
- Claude Code can make all the changes just as if it was installed originally in the cluster, being able to run agent, accessing files and doing inline edits.
Additional Context
_No response_
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗