GitHub Copilot CLI Public Preview Launched Yesterday: AI Coding Agent Now Available In Your Terminal

GitHub is once again breaking new ground in developer productivity, officially opening the public preview of GitHub Copilot CLI—a powerful agentic AI designed to transform the way coding happens in the terminal. With the GitHub Copilot CLI, you can now interact locally and synchronously with an AI that understands both your code and GitHub context, enabling faster, smarter development with no context switching.

GitHub Copilot CLI Public Preview in the Terminal

• Terminal-Native Development
  For the first time, Copilot’s coding agent lives directly in your terminal environment. No more toggling between editors or tools—the AI can plan, build, edit, debug, and refactor code right where you work, in real time.

• Out-of-the-Box GitHub Integration
  With instant, secure authentication using your GitHub account, Copilot CLI accesses your repositories, issues, and pull requests via natural language prompts. Your familiar GitHub context becomes actionable, accelerating common workflows like pull request creation or code review.

• Agentic Capabilities & Full Control
  Copilot CLI is more than just a suggestion tool—it can execute multi-step plans, coordinate complex changes, and preview every action so that nothing runs without your approval. All actions are visible and modifiable before execution, keeping you in control at every step.

• MCP-Powered Extensibility
  Built on the GitHub MCP server, the CLI also supports custom MCP servers. This means developers can extend its core features for specialized or proprietary use cases, supercharging integration and automation options for advanced teams.

Getting Started: Quick Install & Onboarding

Getting up and running with Copilot CLI is simple:

1. Install via npm:

   bash

   npm install -g @github/copilot


2. Authenticate with GitHub:
   Launch copilot in your terminal, then use the /login command to sign in with your GitHub credentials.

3. Start Building:
   Use with any GitHub Copilot plan (Pro, Pro+, Business, or Enterprise). Copilot CLI works across Linux, macOS, and (experimental) native Windows.

• Official Installation Guide

• Copilot CLI Repo & Product Page

• Documentation & Concepts

Model Support and Usage

Copilot CLI runs by default on Claude Sonnet 4, with support for GPT-5 via environment variables, giving power users more choices and flexibility. The tool draws from your Copilot quota, and premium requests are managed according to your plan.

Real-World Use Cases: AI Where Devs Work

Whether exploring and initializing a new repository, implementing feature requests from issues, refactoring, debugging, or just speeding up repetitive DevOps, Copilot CLI acts as an always-available agent.
For example, from a project directory you can ask Copilot:

1. “Explain what changed in this commit.”

2. “Refactor main.py to use async functions.”

3. “Create a new branch, implement feature X, and open a PR.”

It can even execute programmatic commands via prompt, with robust safety options for task approval.

Security and Control

Every action must be explicitly approved before execution—nothing happens without your say-so. This “full control” approach is designed to keep your workflows safe and accountable, even when automating with complex agent functionality.

Feedback and Roadmap

While the public preview is already feature-rich, GitHub invites feedback via /feedback directly in the CLI to drive rapid iteration. New features, model support, and broader third-party extensions are expected to follow. Teams using Copilot Pro, Business, or Enterprise can now explore how the Copilot CLI fits in their daily flows, while the previous gh-copilot extension will retire in October to make way for this expanded toolset.

GitHub Copilot CLI’s public preview marks a major leap in the agentic AI era—inviting developers to collaborate directly with intelligent, secure assistants in the terminal. It blends the world’s best code generation models, GitHub’s trusted platform, and tightly integrated agent workflows to supercharge local and cloud development.
Ready to try Copilot CLI yourself?

1. Get started with Copilot CLI (GitHub Changelog)

2. Install from npm

3. Read GitHub Copilot CLI documentation