Sunday, November 15, 2026 Independent reporting on artificial intelligence Newsletter
AI News Beats
AI for Business

GitHub Copilot app adds general access to Enterprise Server

Enterprise Server repositories, issues and pull requests can now be used in the desktop app, but cloud sessions remain unavailable for those repositories.

An open laptop showing code sits beside a mug and notebook
Photo: Daniil Komov / Pexels

GitHub made Enterprise Server support generally available in its Copilot desktop app on November 13, moving the feature out of an opt-in preview. Developers can connect work hosted on their company’s GitHub Enterprise Server instance, including repositories, issues and pull requests.

The release requires Copilot app version 1.2.0 or later. It does not enable cloud sessions for Enterprise Server. That limitation matters for teams deciding whether an agent will operate on a developer’s machine or in a separately hosted environment.

Repository identity and model access are separate

GitHub Docs’ Enterprise Server sign-in instructions require an initial account on GitHub.com or GHE.com. The user then adds an Enterprise Server account through the app’s Accounts settings and completes authorization in a browser. Connecting the repository account is therefore an additional step, not a replacement for the initial cloud-account sign-in.

To start making requests, the instructions say the user must also be signed in to a cloud account with a Copilot license, or configure their own model provider. Enterprise Server access alone does not establish the model entitlement. The distinction gives administrators two things to verify: access to company repositories and access to the service that will answer coding requests.

Existing organizational controls still apply

GitHub Docs’ team-rollout guide describes a desktop application governed through individual licenses and enterprise or organization policies. The Copilot app policy is separate from the Copilot CLI policy. Enabling one does not enable the other, even when both tools are available to the same developer.

The guide recommends establishing model availability and any required external-tool permissions before a pilot begins. Model Context Protocol servers, which connect an agent to additional tools, can be restricted to approved configurations. Allowing a server through policy does not install it on developers’ computers.

GitHub’s suggested pilot uses a small repository task, such as adding tests for existing behavior, followed by review of the code changes and the normal pull-request process. This is deployment guidance rather than evidence that the new integration improves productivity or produces correct code without review.

Local execution still needs deliberate boundaries

GitHub Docs’ agent-session guide distinguishes the agent’s working mode from its tool permissions. A plan-first workflow controls how work is organized; permissions control whether particular commands or other tool actions need approval. These are separate settings, not interchangeable safeguards.

The same guide describes local sandboxing as restricting the filesystem, network and credential access available to agent-run tools. Separate Git worktrees provide distinct working copies for parallel tasks, reducing interference between file changes. Neither description means that Enterprise Server support provides an offline model or removes the need to inspect a proposed change. The release establishes repository integration; teams still need to validate their permitted models, local environment and review process before broader use.

Stay on top of AI

Get the stories that matter in artificial intelligence, straight to your inbox.

Free. One email with the stories that matter. Unsubscribe any time. See our privacy policy.