Copilot CLI

About the Project

GitHub Copilot CLI is GitHub's official terminal-based AI coding assistant. It supports BYOK (Bring Your Own Key) mode for connecting to custom model endpoints, and can integrate directly with Models Hub through environment variables. It supports Agent mode, tool calling, and MCP.

• Official docs: https://docs.github.com/copilot/how-tos/copilot-cli/cli-getting-started
• Package: https://www.npmjs.com/package/@github/copilot

This guide explains how to point Copilot CLI at Models Hub using BYOK mode, so you can use models distributed through the Models Hub gateway for everyday coding tasks.

📦 Before You Begin

What you'll need

• Node.js 22 or later
• A working Models Hub address (must end with /v1)
• A working Models Hub API Key (generated in the console)
• The name of the model you want to use (such as claude-sonnet-4-6, gemini-2.5-flash, etc.), which must exactly match the model ID in the Models Hub console

🚀 Step 1: Install Copilot CLI

npm install -g @github/copilot

After installation, verify the version:

copilot --version

For detailed installation instructions, see the official getting started guide.

🔧 Step 2: Configure Models Hub (the core step)

Copilot CLI reads custom provider configuration from environment variables. Using the anthropic type is recommended for the best compatibility.

Linux / macOS

export MODELSOK_API_KEY="ms_live_your_key"

Windows (PowerShell)

$env:COPILOT_PROVIDER_TYPE="anthropic"
$env:COPILOT_PROVIDER_BASE_URL="https://modelsok.com/v1"
$env:COPILOT_PROVIDER_API_KEY="your-modelsok-api-key"
$env:COPILOT_MODEL="claude-sonnet-4-6"

Why the anthropic type is recommended

Some models (especially reasoning models with a chain of thought) require reasoning_content to be passed back verbatim in the next request. Copilot CLI's OpenAI integration does not support this mechanism and may trigger a 400 error. Switching to an Anthropic Messages API-compatible endpoint avoids this issue.

If you only use non-reasoning models (such as gemini-2.5-flash), you can also set COPILOT_PROVIDER_TYPE to openai.

Optional: Adjust context and output limits

export MODELSOK_API_KEY="ms_live_your_key"

✅ Step 3: Launch and Verify

copilot

Enter any programming question (for example, "Write me a Python function that reads a JSON file"). If the model responds normally, the configuration is successful.

copilot help providers

This lists all provider-related environment variables and their descriptions.

🔄 Switching Models

Just change the COPILOT_MODEL environment variable and restart copilot:

copilot

The model name must exactly match the model ID exposed in the Models Hub console.

❓ FAQ