GitHub Copilot CLI - Bring Your Own Key (BYOK)

In this blog post, we will explore how to use your own LLMs instead of GitHub hosted models. This is called BYOK - Bring Your Own Key. BYOK helps us to

• Keep your data private: Your prompts and code stay within your own systems.
• Choose what works best: Use the AI model that fits your team’s needs and way of working.
• Save on costs: If you already have systems running for AI workloads, adding a Copilot endpoint can cost very little extra.

In this blog I will be using Azure Models. To use Azure, we need to set the following environment variables.

export COPILOT_PROVIDER_BASE_URL=https://<YOUR_PROJECT_NAME>.services.ai.azure.com/openai/v1
export COPILOT_PROVIDER_TYPE=azure
export COPILOT_PROVIDER_API_KEY=YOUR_AZURE_FOUNDRY_API_KEY
export COPILOT_MODEL=DeepSeek-V4-Pro

Now we can run the Copilot CLI using copilot command - which will use the DeepSeek-V4-Pro model.

And if we want to use Ollama models or any other local LLMs, we need to set the following environment variables.

export COPILOT_PROVIDER_BASE_URL=https://<your-endpoint>/v1
export COPILOT_MODEL=<your-model-name>

This way we can use Azure Foundry models in GitHub Copilot CLI. We can use local models when we are working with Privacy-sensitive code or Offline workflows or learning & experimentation with LLM behavior. When it is not recommended for High-accuracy, large-context, or production-critical tasks - cloud models remain superior in speed and reliability.

Happy Programming