GitHub Copilot Enterprise and GitHub Enterprise Server: What Runs Where?
🔍 WiseChecker

GitHub Copilot Enterprise and GitHub Enterprise Server: What Runs Where?

GitHub Copilot Enterprise is not a self-hosted edition of GitHub Enterprise Server. One is an AI service plan; the other is a customer-managed GitHub deployment. Hosting repositories on your own server does not automatically keep Copilot prompts or inference on that server.

There are also limited, separately configured options for GitHub Enterprise Server users, including a Copilot CLI technical preview. Distinguish those options before choosing an architecture, approving confidential code use or estimating costs.

ADVERTISEMENT

1. Separate the hosting product from the Copilot plan

GitHub Enterprise Cloud is GitHub’s hosted enterprise platform. GitHub Enterprise Server, or GHES, is the customer-managed platform. Copilot Business and Copilot Enterprise are separate AI plans; GitHub describes Copilot Enterprise as intended for enterprises using Enterprise Cloud. See the Copilot plan reference.

Do not plan for a “Copilot Enterprise Server” appliance with a bundled local model, a fixed extra CPU/RAM requirement or feature parity with cloud Copilot. Those are not the products described in the current official documentation.

The general plans page still says Copilot is unavailable for GHES, while the dedicated GHES 3.22 guide explains narrower ways to use it. For a GHES deployment, read that version-specific guide rather than interpreting either statement as full support for every cloud feature or as a ban on every local client.

2. Understand cloud-licensed use with a GHES codebase

The GHES guide describes using an enterprise account on GitHub.com or GHE.com to assign Copilot licenses and configure policies. Developers authenticate to that cloud account in supported local clients. Copilot can work with code checked out from the server; supported GitHub tooling may also connect to the GHES instance.

This does not move the entire cloud Copilot experience into the GHES website. Nor does a private repository URL prove that model requests stay inside your network. Evaluate the client, model and enabled tools as part of the data flow.

For cloud-connected use, administrators should review the Copilot network requirements and approve the necessary endpoints. Do not copy cloud tokens onto the GHES appliance or bypass a restricted network to obtain suggestions.

ADVERTISEMENT

3. Treat offline Copilot CLI as a separate technical preview

GHES 3.22 documentation describes a Copilot CLI technical preview for disconnected or air-gapped configurations. An administrator configures a model provider for the server, and users connect the CLI with their GHES credentials. This route does not require a Copilot license, but it still needs a suitable model provider and an approved deployment.

The setup guide specifies an upstream model endpoint and provider credentials. That endpoint may be external or internally hosted. Therefore, “no GitHub cloud connection” is not the same as “no external data processing.” A genuinely isolated design must also keep the selected model endpoint and permitted tools within the approved boundary.

This is not a bundled GitHub model or a promise of every cloud feature in every IDE. Features that rely on GitHub cloud connectivity are unavailable in the offline configuration. Confirm the exact GHES version, supported client, provider and preview limitations with the administrator before adoption.

4. Review residency, privacy and feature controls separately

Cloud users with regional requirements should review GitHub Copilot with data residency. The documented policy restricts processing to supported regional endpoints and constrains model availability. It is disabled by default and affects usage cost. Regional cloud processing is still cloud processing, not inference on a company-owned server.

For any deployment, ask the security team to verify the actual provider, data retention terms, logging, allowed repositories and tool permissions. Do not infer a certification or regulatory approval from the word “Enterprise.” Check the contract and applicable compliance evidence for the specific service and region.

Feature controls also depend on the plan and client. A model selector, pull-request workflow or code-review feature available in one surface may not exist in another. Use the Copilot product documentation to scope the proposed workflow, then test it with non-sensitive data.

ADVERTISEMENT

5. Compare costs and choose an approved architecture

As checked on October 14, 2026, GitHub lists Copilot Business at $19 USD and Copilot Enterprise at $39 USD per granted seat per month, with AI-credit allowances and additional usage rules. Confirm current terms in the official plan reference or your contract; a seat price is not an unlimited all-model usage promise.

For the CLI provider route, budget for the chosen provider or local inference infrastructure, operations and the underlying GitHub deployment. Do not apply an invented $468 annual “server Copilot” price or a 50-seat minimum to it.

  • Need the hosted enterprise Copilot workflow: evaluate the cloud plan, policy and client combination.
  • Keep repositories on GHES but permit approved cloud inference: evaluate cloud-assigned access for the required local clients.
  • Require disconnected execution: evaluate the version-specific CLI technical preview and an approved internal provider; do not assume full cloud parity.

Record the selected architecture, its data routes and its unsupported features before procurement or rollout. That is a more reliable decision than comparing two supposedly equivalent “cloud and server” Copilot editions.

Official documentation checked October 14, 2026. GHES technical-preview availability and contract requirements must be rechecked before deployment.

ADVERTISEMENT

Did this guide solve your problem?