How to Configure GitHub Copilot CLI to Use Z.ai's GLM Coding Plan

Mon, 20 Apr 2026 00:00:00 GMT

Introduction

GitHub Copilot CLI recently added support for BYOK (Bring Your Own Key). What this means is that you can configure Copilot CLI to use an Anthropic or OpenAI-compatible endpoint and use a model without needing a GitHub Copilot subscription. In this post I'll explain how to use the GLM-5.1 model from Z.ai's GLM Coding Plan subscription.

But first, here is a short explanation of the tools used.

GitHub Copilot CLI is the official TUI AI agent by GitHub. It's one of the many ways GitHub provides to use their Copilot subscription and integrates nicely with Visual Studio Code as well.

Z.ai's GLM Coding Plan is Z.ai's subscription built specifically for AI-powered coding. They have documentation on how to use the subscription from many different coding agents.

GLM-5.1 is Z.ai's flagship model (as of writing this). It's competitive with the flagship models provided by other major AI labs.

Prerequisites

Linux or macOS
GitHub Copilot CLI installed
An active Z.ai Coding subscription
- A Z.ai API key

Configuration

There are two ways to configure Copilot CLI: using the Anthropic-compatible API or using the OpenAI-compatible API. I haven't found any difference between those two integration methods.

Unfortunately there is no way to configure this in the ~/.copilot/config.json file. Instead you need to use the appropriate environment variables. You probably want to save the commands below as a shell alias, if you are going to use them often.

A. Configure using the OpenAI-compatible API

COPILOT_OFFLINE=true \
COPILOT_PROVIDER_TYPE=openai \
COPILOT_PROVIDER_BASE_URL=https://api.z.ai/api/coding/paas/v4 \
COPILOT_PROVIDER_API_KEY=REPLACE_WITH_YOUR_ZAI_KEY \
COPILOT_MODEL=glm-5.1 \
COPILOT_PROVIDER_MAX_PROMPT_TOKENS=204800 \
COPILOT_PROVIDER_MAX_OUTPUT_TOKENS=131072 \
copilot

B. Configure using the Anthropic-compatible API