> ## Documentation Index
> Fetch the complete documentation index at: https://docs.gravitex.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# GitHub Copilot App

> Connect the GitHub Copilot desktop client to GravitexAI GPT and Claude models

## 1. Overview

GitHub Copilot App is GitHub's desktop coding Agent. Add a local project or GitHub repository and use natural-language instructions for coding tasks. **BYOK (bring your own provider)** lets you select GravitexAI GPT or Claude models.

| Capability | Description |
| - | - |
| GUI setup | Enter the URL and API key in Model providers and add models without editing files |
| Multiple models | Configure separate GPT and Claude providers and select a model per chat |
| Adjustable autonomy | Interactive, Plan, and Autopilot session modes |
| Usage-based billing | Custom-provider sessions are billed to GravitexAI, without consuming Copilot allowance |

> [!NOTE]
>
> Applies to **GitHub Copilot App 1.1.x (October 2026)**. BYOK is in **Public Preview**, so interface wording may change. The screenshots use the English interface.

## 2. System requirements

* macOS or Windows.
* [Git](https://git-scm.com/) installed.
* Access to GitHub for sign-in and to the GravitexAI API.

## 3. Prerequisites

* A GitHub account for signing in.
* An API key (`sk-xxxxxxxxxx`) from the [GravitexAI Console](https://maas.gravitex.ai/#/api-keys).

> [!NOTE]
>
> A paid GitHub Copilot subscription is not required when using only your own model provider.

## 4. Quick start

### Step 1: Install the client

> [tab] macOS
>
> Install with Homebrew:
>
> ```bash theme={null}
> brew install --cask github-copilot-app
> ```
>
> Or download a package for your processor:
>
> * [Apple Silicon](https://github.com/github/app/releases/latest/download/GitHub-Copilot-darwin-arm64.dmg)
> * [Intel](https://github.com/github/app/releases/latest/download/GitHub-Copilot-darwin-x64.dmg)
>
> Open the DMG, drag GitHub Copilot to Applications, and launch it.

> [tab] Windows
>
> Download and run the matching installer:
>
> * [Windows x64](https://github.com/github/app/releases/latest/download/GitHub-Copilot-windows-x64-setup.exe)
> * [Windows ARM64](https://github.com/github/app/releases/latest/download/GitHub-Copilot-windows-arm64-setup.exe)

Other versions are available on [GitHub Releases](https://github.com/github/app/releases).

### Step 2: Sign in

Click **Sign in to GitHub** and authorize in your browser.

* **Without a Copilot subscription:** choose **continue with your own model provider** during onboarding, enter the values from step 4, and click **Save and continue**.
* **With a subscription:** enter the main interface and add a provider through Settings.

### Step 3: Open provider settings

| System | Open Settings |
| - | - |
| macOS | **GitHub Copilot → Settings**, or `⌘ ,` |
| Windows | **Settings** in the avatar/menu area, or `Ctrl + ,` |

Select **Model providers** on the left, then **Add provider**.

![Open Model providers and click Add provider](https://mintcdn.com/gravitexai/QXITE-rsvJBzaoaP/images/github-copilot-app-1.png?fit=max\&auto=format\&n=QXITE-rsvJBzaoaP\&q=85\&s=22a7af4e2a7f430c5055ab78be1e12e9)

### Step 4: Enter GravitexAI connection details

Add GPT or Claude as needed. To use both, create two separate providers; the same GravitexAI API key can be used for each.

| Field | GPT provider | Claude provider |
| - | - | - |
| **Provider type** | **Any OpenAI-compatible HTTP endpoint** | **Anthropic** |
| **Display name** | `Gravitex GPT` | `Gravitex Claude` |
| **Base URL** | `https://api.gravitex.ai/v1` | `https://api.gravitex.ai` |
| **API key** | Your GravitexAI API key | The same API key |
| **Wire API / API format**, if shown | **Responses**, using `/v1/responses` | No setting needed; uses `/v1/messages` |

Click **Add provider** to save a new provider, or **Save changes** when editing one.

> [!WARNING]
>
> GPT requires an OpenAI-compatible provider, a Base URL **with `/v1`**, and Responses when the format option is available. Claude requires **Anthropic** and a Base URL **without `/v1`**. Claude does not support `/v1/responses`; do not add it to the GPT provider.

The example uses the display name `Anthropic`; you can rename it `Gravitex Claude` for easier identification. Leave **Custom headers (JSON)** empty unless specifically needed.

![Anthropic configuration without /v1 in the Base URL](https://mintcdn.com/gravitexai/QXITE-rsvJBzaoaP/images/github-copilot-app-2.png?fit=max\&auto=format\&n=QXITE-rsvJBzaoaP\&q=85\&s=1f7b5a0498201a44f3e79152a642818a)

### Step 5: Add models

Select models from the catalog after saving, or enter model IDs manually.

| Provider | Example model IDs |
| - | - |
| **Gravitex GPT** | `gpt-5.5`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.3-codex`, `gpt-6-sol` |
| **Gravitex Claude** | `claude-opus-5-5`, `claude-sonnet-5-5`, `claude-haiku-5-5`, `claude-fable-5-1` |

> [!TIP]
>
> Use exact IDs from the [GravitexAI Model Catalog](https://maas.gravitex.ai/#/api-models), subject to account availability. Save the API key before browsing the catalog.

### Step 6: Add a project and select a model

1. Click **+** beside **Projects** to add a project:

   | Option | Source |
   | - | - |
   | **Local folder or repository** | A folder or repository on your computer |
   | **GitHub repository** | A repository hosted on GitHub |
   | **Repository URL** | A repository address |

2. Create a **New session**.

3. Open the model picker below the composer and select a model under **Gravitex GPT** or **Gravitex Claude**. Adjust **Reasoning effort** if available.

4. Send a test message and confirm a normal response.

![Select the configured Claude model in a session](https://mintcdn.com/gravitexai/QXITE-rsvJBzaoaP/images/github-copilot-app-3.png?fit=max\&auto=format\&n=QXITE-rsvJBzaoaP\&q=85\&s=b7ac69928e0ba21e80c44cd5887debb2)

> [!NOTE]
>
> The screenshot keeps the default provider name `Anthropic`. If you changed Display name, your chosen name appears beside the model instead.

## 5. Session modes

| Mode | Behavior |
| - | - |
| **Interactive** | Default: suggests changes and waits for confirmation |
| **Plan** | Prepares a plan for review before executing |
| **Autopilot** | Changes code, runs commands and tests, and continues iterating |

> [!WARNING]
>
> Autopilot directly modifies files and runs commands. Use it in projects tracked with Git and review the resulting changes.

## 6. Success checks

* A provider is saved in Model providers with at least one compatible model.
* The configured model appears and can be selected in a new session.
* Messages receive replies without `invalid api key`, `endpoint_not_found`, or model-availability errors.

> [!NOTE]
>
> Custom-provider usage is billed to GravitexAI and does not consume Copilot allowance. Rate limits do not automatically switch these sessions to Copilot's Auto model.

## 7. Troubleshooting

> [accordion] Claude returns 404 / endpoint\_not\_found
>
> Use Anthropic with [https://api.gravitex.ai](https://api.gravitex.ai), without /v1. A path such as /v1/v1/messages means /v1 was duplicated. Do not configure Claude as an OpenAI-compatible Responses provider.

> [accordion] GPT returns 404
>
> Use Any OpenAI-compatible HTTP endpoint with [https://api.gravitex.ai/v1](https://api.gravitex.ai/v1). Choose Responses when a format option is available.

> [accordion] Invalid API key for this provider
>
> Edit the provider in Model providers, enter the complete key without extra spaces, and save. Check key validity in the GravitexAI Console.

> [accordion] The provider does not expose a /models endpoint
>
> For `Provider doesn't expose a /models endpoint at this base URL`, verify the type and URL: GPT uses [https://api.gravitex.ai/v1](https://api.gravitex.ai/v1); Claude uses [https://api.gravitex.ai](https://api.gravitex.ai). If catalog browsing still fails, try entering a model ID manually.

> [accordion] The catalog asks for an API key first
>
> For `Add an API key for this provider before browsing the catalog`, save the provider's key before opening its catalog.

> [accordion] The model picker does not show GravitexAI models
>
> Check that the provider is saved and models are selected or manually added. Display names are labels only; look under Anthropic if that default name was retained. Restart the app if needed.

> [accordion] Where is the API key stored?
>
> In the system credential store: macOS Keychain or Windows Credential Manager. Do not expose keys in public screenshots, chats, or repositories.

### Optional API checks

Use these requests to distinguish client configuration problems from API problems. Replace the placeholder key and use a model available to your account.

```bash GPT（Responses） theme={null}
curl https://api.gravitex.ai/v1/responses \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.5","input":"say OK"}'
```

```bash Claude（Messages） theme={null}
curl https://api.gravitex.ai/v1/messages \
  -H "x-api-key: sk-your-api-key" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-sonnet-5-5","max_tokens":20,"messages":[{"role":"user","content":"say OK"}]}'
```

> [!NOTE]
>
> In Windows PowerShell, use `curl.exe` and replace each trailing `\` with a backtick `` ` ``.

## 8. Resources

* [GitHub Copilot App Releases](https://github.com/github/app/releases)
* [GravitexAI API keys](https://maas.gravitex.ai/#/api-keys)
* [GravitexAI models](https://maas.gravitex.ai/#/api-models)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.