Your company’s code lives on GitHub Enterprise Server, but your AI coding workspace is elsewhere? GitHub’s Copilot desktop app can now connect to that self-hosted work.
On November 13, 2026, GitHub announced that GitHub Enterprise Server (GHES) account support in the Copilot app had moved from opt-in preview to general availability. The practical win: access remote repositories, issues, and pull requests from the same app instead of treating your enterprise instance as a separate world.
How to connect your GHES account
GitHub’s documented setup is short:
- Install or update the GitHub Copilot app to version 1.2.0 or later.
- Sign in to the app with a GitHub.com or GHE.com account first.
- Open Settings > Accounts > Add account.
- Choose GitHub Enterprise Server, then complete browser authorization for your enterprise instance.
Once connected, you can work with repositories, pull requests, and issues hosted on your GHES instance. This is account integration, not a command-line migration or a requirement to move your repositories to GitHub.com.
The catch: cloud sessions aren’t supported
GitHub explicitly says cloud sessions are unavailable for GHES in this integration. Do not assume that every agent workflow available for cloud-hosted GitHub repositories will work on your private server.
There’s also a separate AI access requirement. GitHub’s documentation says that to make requests, you need to be signed into a cloud account with a Copilot license, or configure your own model provider. Simply adding a GHES account does not automatically supply a model or entitlement.
For enterprise teams, check your organization’s authentication, network access, and data-handling rules before connecting a coding assistant to sensitive repositories. The release announcement does not promise that repository content stays entirely on your GHES server.
Quick takeaway
If you already use GHES, update the Copilot app, add the enterprise account, and try a repository or pull-request workflow first. Confirm the authentication and model setup before expecting agent sessions to work.
