Article

Connect GitHub Enterprise Server to the Copilot app

A setup guide for teams with self-hosted GitHub repositories: connect accounts, choose model access and check costs, policies and session limits.

Editorial illustration for Connect GitHub Enterprise Server to the Copilot app: code brackets represent developer tools. Not documentary evidence.

Teams using GitHub Enterprise Server (GHES), GitHub hosted on their own server, can connect those repositories to the Copilot desktop app. Use app version 1.2.0 or later, sign in with a GitHub.com or GHE.com account first, then add the server account; GitHub made this integration generally available on November 13, 2026.

To make AI requests, you also need a Copilot entitlement on the cloud account or a configured model provider. Adding the server account gives the app repository access; it does not by itself supply a model. Cloud sessions are not available for GHES, so plan to run the session on your computer.

Connect the cloud and server accounts

Install Git before starting. GitHub’s app documentation lists macOS, Linux and Windows support, and its quickstart explains installation and onboarding. The 1.2.0 requirement is the app version, not a GHES server release; the announcement does not specify a minimum server version.

  1. Open the app and complete the initial cloud sign-in. For a GHE.com account, select GitHub Enterprise Cloud and enter your cloud hostname when prompted.

  2. In the app, open Settings, select Accounts, then choose Add account and GitHub Enterprise Server.

  3. Complete the browser authorization prompts. GitHub’s server sign-in instructions describe this as an additional account, not a replacement for the initial cloud sign-in.

For an employer-provided Copilot licence, use the cloud identity assigned that licence. Licensing and administration happen on GitHub.com or GHE.com; this integration does not put Copilot Chat inside the GHES website. Your repositories can remain on the server.

Choose how the app gets its models

Use your Copilot plan. Keep the licensed cloud account signed in. All Copilot plans include app access, but model choices, allowances and workplace controls differ. Copilot Free is limited and uses automatic model selection; it is not unlimited team access.

Use your own provider. GitHub’s provider setup guide says this option needs GitHub sign-in but no Copilot plan. In Settings, select Model providers, then Add provider; choose the provider and supply its required details in the app. The models then become selectable for a session. Supported options include Anthropic, Microsoft Foundry, Ollama and LM Studio. This feature remains in public preview, even though GHES account integration is generally available.

A separate provider may charge for usage; running a local model also requires suitable computing resources. “No Copilot plan” does not mean no operating cost. With user-configured providers, model requests go directly to that provider. Centrally configured enterprise models are a different arrangement: requests pass through GitHub, and users need a licence and internet access.

For an existing rollout, check saved model choices against the November 16 Copilot model retirements. For a rollout extending into next month, the December 14 model-retirement guide covers the next announced changes.

Check the app policy before rolling it out

For Business and Enterprise users, the dedicated Copilot app policy must be enabled. GitHub enables it by default, but it is separate from permission to use Copilot CLI.

Enterprise owners can find the client controls under AI controls → Copilot → Configure features & clients. Organization owners manage feature and model policies under Settings → Copilot; they cannot override an enforced enterprise setting. Confirm both the app and the intended model are allowed before distributing onboarding instructions.

Estimate the team’s Copilot bill

GitHub’s published USD plan rates list Copilot Business at $19 per assigned seat per month and Copilot Enterprise at $39. Business contributes 1,900 monthly AI credits per user; Enterprise contributes 3,900. These credits form a shared pool at the billing-entity level, rather than separate developer allowances, and unused credits expire monthly.

GitHub values one AI credit at $0.01. Model usage depends on the amount of input, output and cached context processed, measured in tokens. A credit allowance therefore cannot be translated into a fixed number of coding tasks.

Illustrative five-person Business budget: Assume five seats for a full month and hypothetical aggregate consumption of 12,500 credits, with paid usage permitted.

  • Seat charges: 5 × $19 = $95.

  • Included pool: 5 × 1,900 = 9,500 credits.

  • Usage beyond the pool: (12,500 − 9,500) × $0.01 = $30.

  • Total Copilot charge: $125 for the month.

This calculation applies the published allowances and credit rate, not measured usage. It assumes no other seats contribute to the pool and no control blocks consumption. It excludes tax, discounts, mid-month seat changes, existing GitHub Enterprise charges and separately billed providers.

Additional paid usage is enabled by default on Business and Enterprise. To stay within the included pool, an administrator must disable AI credits paid usage. Do not substitute a $0 user-level budget: GitHub’s budget rules say that blocks credit-consuming features even when pooled credits remain. A user cap limits total consumption; an enforced spending limit controls extra charges.

Some enterprises have a different option called Copilot access: a $0 Copilot seat charge for Enterprise Cloud-licensed users, or $10 per user monthly for Copilot Standalone. Both have no included credits and bill metered usage from the start. As of November 16, eligibility requires working with a GitHub or Microsoft account team; self-serve customers are not yet eligible. Switching is voluntary, seats still need assigning, and this model cannot coexist with Business or Enterprise seats in the same enterprise. A GHES login alone does not establish eligibility.

Start a local session and review the change

After authorization and model setup, connect the intended repository or open its existing local checkout. Select that repository when starting a session. GitHub’s session controls distinguish the execution environment from the working copy: choose This computer for GHES; New worktree creates a separate working copy, while Current checkout uses the existing one.

A sensible first task is a small change your team can review. Consider Interactive or Plan mode and deliberately choose the tool-permission setting; session mode and tool approval are separate controls. After the change, inspect Changes, run the project’s relevant checks and use Create PR to request review.

Local execution does not automatically mean local AI processing: your selected provider determines where model requests go. The documented app setup still starts with cloud sign-in, so this is not a fully disconnected onboarding recipe.

If a repository is visible but AI requests fail, a useful troubleshooting order is to check the cloud entitlement or provider configuration, then effective model policy, then available budget. These are separate requirements in GitHub’s sign-in, policy and billing documentation. Reauthorizing the server account does not resolve every model-access problem.