copilot

Copilot CLI Adds BYOK and Local Model Support

Copilot CLI now supports bring-your-own-key models, local inference, and offline mode. Use your own providers, run air-gapped, skip GitHub auth.

TL;DR

Copilot CLI now supports bring-your-own-key (BYOK) — use Azure OpenAI, Anthropic, or any OpenAI-compatible endpoint
Run fully local models (Ollama, vLLM, Foundry Local) in air-gapped environments with offline mode
GitHub authentication is now optional when using your own provider

BYOK model provider support — Configure Copilot CLI to use Azure OpenAI, Anthropic, or any OpenAI-compatible endpoint via environment variables. Works with remote services and locally running models.
Offline mode — Set COPILOT_OFFLINE=true to disable all GitHub server communication and telemetry. Enables fully air-gapped development workflows when paired with a local model.
Optional GitHub authentication — Use Copilot CLI with just your provider credentials. GitHub sign-in is now optional but still unlocks features like /delegate, Code Search, and the GitHub MCP server.

Your model must support tool calling and streaming. Aim for at least 128k token context window for best results.
Built-in sub-agents (explore, task, code-review) automatically inherit your provider configuration.
Invalid provider configs trigger actionable error messages — no silent fallback to GitHub-hosted models.
Run copilot help providers in the terminal for quick setup instructions.

Update Copilot CLI via gh copilot update or see the setup docs for full configuration details.