GitHub Copilot App

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

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 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.

A paid GitHub Copilot subscription is not required when using only your own model provider.

4. Quick start

Step 1: Install the client

macOS
Windows

Install with Homebrew:

brew install --cask github-copilot-app

Or download a package for your processor:

Open the DMG, drag GitHub Copilot to Applications, and launch it.

Other versions are available on GitHub Releases. 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

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.

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

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`

Use exact IDs from the GravitexAI Model Catalog, subject to account availability. Save the API key before browsing the catalog.

Step 6: Add a project and select a model

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
Create a New session.
Open the model picker below the composer and select a model under Gravitex GPT or Gravitex Claude. Adjust Reasoning effort if available.
Send a test message and confirm a normal response.

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

Select the configured Claude model in a session

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

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.

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

Claude returns 404 / endpoint_not_found

Use Anthropic with 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.

GPT returns 404

Use Any OpenAI-compatible HTTP endpoint with https://api.gravitex.ai/v1. Choose Responses when a format option is available.

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.

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; Claude uses https://api.gravitex.ai. If catalog browsing still fails, try entering a model ID manually.

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.

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.

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.

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"}'

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"}]}'

In Windows PowerShell, use curl.exe and replace each trailing \ with a backtick `.

Chat AI

Coding

AI Agent

Tech Engineering

Translation

1. Overview

2. System requirements

3. Prerequisites

4. Quick start

Step 1: Install the client

Step 3: Open provider settings

Step 4: Enter GravitexAI connection details

Step 5: Add models

Step 6: Add a project and select a model

5. Session modes

6. Success checks

7. Troubleshooting

Optional API checks

8. Resources

​1. Overview

​2. System requirements

​3. Prerequisites

​4. Quick start

​Step 1: Install the client

​Step 2: Sign in

​Step 3: Open provider settings

​Step 4: Enter GravitexAI connection details

​Step 5: Add models

​Step 6: Add a project and select a model

​5. Session modes

​6. Success checks

​7. Troubleshooting

​Optional API checks

​8. Resources

1. Overview

2. System requirements

3. Prerequisites

4. Quick start

Step 1: Install the client

Step 2: Sign in

Step 3: Open provider settings

Step 4: Enter GravitexAI connection details

Step 5: Add models

Step 6: Add a project and select a model

5. Session modes

6. Success checks

7. Troubleshooting

Optional API checks

8. Resources