How to sign-in and use GitHub Copilot in an air-gapped environment?

[doggy8088]

Select Topic Area

Question

Copilot Feature Area

VS Code

Body

I have deployed some GPU resources in an air-gapped environment and installed some models using vLLM. I hope to call them directly using VS Code + GitHub Copilot, but I am currently stuck because I cannot log in to GitHub, which prevents me from using GitHub Copilot. Are there any solutions?

[saad2134]

GitHub Copilot won’t work offline, iit requires GitHub login + cloud.

For an air-gapped setup with vLLM, use a local VS Code extension (e.g. Continue.dev, CodeGPT, or a custom LSP) that connects directly to your models. That’s your path to Copilot-like help without internet.

[doggy8088]

I mean. I have GPU but I still can't sign-in GitHub Copilot in air-gapped environment so that I still can't use GitHub Copilot that connect to my vLLM.

[Kumarageay-ds]

Use an on-prem / self-hosted LLM alternative

Since you already installed models with vLLM, you can integrate them with VS Code through extensions like:

Continue.dev
(VS Code extension that lets you connect to local or self-hosted LLMs).

CodeGPT
or similar.

These let you point the IDE at your local model API instead of GitHub Copilot’s cloud endpoint.

Enterprise GitHub Copilot (with proxy)

If your company uses GitHub Copilot Enterprise, you can set up a secure proxy or bastion host that authenticates with GitHub outside the air-gapped zone, then routes completions internally.

This requires IT/security buy-in and usually isn’t possible in a fully air-gapped environment.

Manual offline usage

Run your own completion server (via vLLM + OpenAI-compatible API).

Install a VS Code extension (Continue, Tabby, or custom) to call your local endpoint.

This replicates the “Copilot-like” experience without GitHub login.

[henrikcaesar]

Found this link where it look like it is possible to use local/self hosted models:
https://peterfalkingham.com/2025/10/27/using-local-ai-llm-in-vs-code-without-third-party-software/

The problem is to get past the login...

[doggy8088]

Exactly! I have GPU but I still can't sign-in GitHub Copilot in air-gapped environment so that I still can't use GitHub Copilot that connect to my vLLM.

[Sakith-9900]

To use GitHub Copilot in an air-gapped environment, you would need internet access for authentication, as it requires communication with GitHub's servers. One solution is to set up a proxy that allows VS Code to authenticate while maintaining security. If a proxy isn't feasible, you can use offline alternatives like IntelliCode or other local AI models for code completion. Another option is to temporarily connect to the internet just for the authentication process and then disconnect once it's set up. Alternatively, you could use locally deployed models with tools like Tabnine or Kite for code completion, which do not require internet access. Since GitHub Copilot relies on an internet connection, these alternatives could be necessary to continue development in an air-gapped environment.

[doggy8088]

I have a GPU set up on my local machine, and even though GitHub Copilot Chat is an open-source project, I still have to sign in to GitHub to use GitHub Copilot. That’s what’s bothering me right now!

[henrikcaesar]

Found this issue for it microsoft/vscode#246551