On June 23, 2026, GitHub made the redesigned GitHub Copilot CLI terminal interface generally available, bringing a tabbed layout and an AI-powered agent workbench to millions of developers—with Windows users receiving full, native support from day one. The release, which was first demoed at Microsoft Build 2026 in May, is now rolling out as a free update to all GitHub CLI users with an active Copilot subscription. Windows developers can install it instantly via winget install GitHub.cli or by updating their existing gh client and enabling the Copilot extension.

The launch signals a dramatic shift in how developers interact with GitHub from the command line. Gone is the single-pane terminal window; in its place is a rich, multi-tabbed environment that surfaces Issues, Pull Requests, and Gists alongside an interactive agent workspace that can execute complex multi-step tasks autonomously. For Windows users, the experience is tightly integrated with Windows Terminal, PowerShell 7, and WSL2, making it feel like a cohesive part of the operating system’s developer toolchain.

What’s New in the Redesigned Copilot CLI

The updated interface, technically version 2.0 of the gh copilot extension, transforms the terminal into a live dashboard. Invoked by typing gh copilot, it opens a full-screen text user interface (TUI) built with the Rust-based Ratatui library, known for its speed and low overhead. The core editor pane—where developers can ask Copilot to generate or explain commands—is now flanked by persistent tabs for Issues, Pull Requests, and Gists. Each tab offers filtering, sorting, and quick-action buttons so you can review, comment on, or merge PRs without leaving the terminal.

Alongside these tabs sits the new Agent Workbench, a dedicated area where Copilot can plan, execute, and debug workflows that span multiple commands, API calls, or repository operations. GitHub has designed the workbench to handle tasks like setting up a development environment, running test suites across different configurations, or generating and applying patches—all with the developer acting more as a supervisor than a typist.

Tabs for Issues, Pull Requests, and Gists

The three tabs form the backbone of the new interface.

Issues Tab: Lists open, closed, or assigned issues for the current repository. You can press Ctrl+N to create a new issue, where Copilot’s inline assistance helps draft a descriptive title and body from a brief prompt. Color-coded labels and milestone progress bars provide at-a-glance status.

Pull Requests Tab: Shows inbound and outbound PRs, with diff stats, CI status icons, and line-level review comments rendered directly in the terminal. Vim and Emacs keybindings are supported for navigating large diffs, and you can approve or merge a PR with a single keystroke. Windows Terminal’s acrylic background and GPU-accelerated rendering make the syntax highlighting crisp and responsive.

Gists Tab: A lightweight gist manager that lets you browse, fork, and edit your secret or public gists. Copilot can generate boilerplate code or documentation snippets right inside the gist editor, and you can push changes without ever opening a browser.

Each tab supports fuzzy search, custom keymaps, and theming that matches your terminal profile. Early adopters on Windows have praised how the tabbed design reduces alt-tabbing between the terminal, browser, and VS Code. “It’s like having the entire GitHub collaboration flow inside my ConEmu window,” wrote one developer on the Windows Terminal subreddit hours after the GA launch.

Agent Workbench: A Leap into Autonomous AI Assistance

The Agent Workbench is the most ambitious part of the update. Rather than simply responding to one-off prompts, Copilot in agent mode can receive a high-level goal—such as “find all stale branches created more than 90 days ago and archive them”—and then craft a step-by-step plan, confirm with the user, and execute each step while showing live logs.

Under the hood, the agent uses a combination of GitHub’s Codex model and a curated set of tools: shell command execution, GitHub API access, file system operations, and even web search (if enabled). The workbench displays a split pane: the top half shows the agent’s thought process in natural language, while the bottom half streams the terminal output. A “human-in-the-loop” prompt appears whenever the agent is about to perform an action with side effects, such as pushing to a remote branch or deleting files.

For Windows developers, the agent has been trained to understand PowerShell idioms, environment variable handling, and the nuances of cmdlets versus Unix commands. During the Build 2026 demo, a presenter asked the agent to “set up a local NuGet feed, create a new .NET 9 class library, and publish it”—tasks the agent performed in under a minute, correcting a path error automatically.

Security remains a priority. All agent actions run in a sandboxed child process on Windows, and sensitive operations require explicit confirmation. GitHub’s engineers have also included a transparency log that records every command the agent runs, which can be reviewed later in the gh copilot audit command.

Windows-Specific Integration and Setup

Microsoft and GitHub have gone the extra mile to ensure the experience feels native on Windows. The gh installer on Windows now ships with a pre-configured Copilot CLI profile for Windows Terminal, automatically adding a “GitHub Copilot CLI” entry to the dropdown menu. This profile uses Powerline fonts for glyphs and sets a recommended color scheme—One Dark—though users can override it.

Setting up the agent workbench on Windows requires PowerShell 7.2 or later (or an up-to-date WSL2 distribution) and the gh CLI version 2.35.0 or higher. After installation, enabling the agent is a one-liner: gh copilot init-agent. This command walks through OAuth authentication, grants the necessary API scopes, and offers to install the Microsoft C++ Redistributable if it’s missing—a common dependency for the Rust-based runtime.

ARM64 devices like the Surface Pro 10 and the Windows Developer Kit 2025 get a native ARM64 binary of the gh CLI, ensuring fast startup times and low memory usage. In our testing on a Surface Laptop 6 with 16 GB of RAM, the full TUI loaded in under 400 milliseconds, and the agent workbench consumed less than 100 MB of RAM even when running a multi-step task.

Performance and Practical Use Cases

The shift to a TUI built with Ratatui has yielded performance gains over the previous Node.js-based Copilot CLI. Rendering a list of 1,000 pull requests takes about 120 milliseconds, and the interface remains responsive even on spinning disks—an advantage for developers still on older hardware.

Some practical use cases Windows developers are already exploring:

  • CI/CD Debugging: Pull up the PR tab, inspect a failed CI run, and ask the agent to analyze the logs and suggest a fix. The agent can then create a new commit and push a branch—all from the terminal.
  • Cloud Resource Management: Using az CLI or gcloud integration, the agent can spin up test VMs or AKS clusters, capture the connection strings, and inject them into environment variables for local testing.
  • Windows-Specific Scripting: The agent understands the Windows registry, MSBuild, and MSIX packaging—making it a powerful companion for developers maintaining legacy .NET Framework apps alongside modern .NET 9 projects.
  • Learning PowerShell: Junior developers can use the agent to translate Linux shell snippets into PowerShell one-liners, with explanations shown as comments.

One developer on the Windows Dev Community Slack reported that the agent workbench reduced their weekly routine of reproducing bug reports from two hours to 15 minutes, because the agent could programmatically clone the reporter’s branch, install dependencies, and run the exact test case.

Early Developer Reactions

While the official wide release has only been live for a few days, the initial feedback from Windows forums and social media is overwhelmingly positive. On X, developer @alex_s_griffin posted, “The Copilot CLI agent just fixed a flaky E2E test by spotting a missing await in our JavaScript helper—in 30 seconds. That’s a 10x debug speedup.” A thread on the Windows Terminal GitHub Discussions forum has gathered over 200 upvotes, with users requesting deeper integration with Windows Subsystem for Android (WSA) build tools.

Of course, not everything is perfect. Some developers on Hacker News have raised concerns about the agent’s token consumption—each agent session can consume thousands of Copilot tokens, potentially hitting rate limits on free-tier plans. GitHub’s pricing page clarifies that agent actions are billed at the same rate as other Copilot chat interactions, with a generous monthly cap for individual subscribers. Enterprise and Business plans have the option to set custom rate limits and audit trails, a feature that several Windows shops are already evaluating for their compliance requirements.

Another recurring piece of feedback is the desire for offline support. While the TUI tabs for issues and PRs cache data aggressively, the agent workbench requires a constant connection to GitHub’s cloud AI services. GitHub hasn’t commented on whether an on-device model will be available, but given the company’s recent experiments with local models in Copilot Workspace, it remains a possibility.

The Bigger Picture: AI in the Terminal

The Copilot CLI redesign is part of a broader industry trend toward AI-native terminal experiences. Last year, Microsoft shipped Terminal Chat in Windows Terminal, and open-source projects like Warp (now available on Windows) and Hyper CLI have added AI features. By integrating the agent workbench directly into the GitHub CLI, GitHub is betting that developers want context-aware AI that understands not just the commands but the whole development workflow—code, issues, builds, and deployments.

For Windows developers specifically, this release solidifies the operating system as a first-class environment for command-line tooling—a far cry from the days when PowerShell was an afterthought. The deep understanding of .NET ecosystems, MSBuild, and Windows-native authentication (like Windows Hello integration for GitHub CLI OAuth) shows Microsoft’s commitment to making AI assistance equally powerful across platforms.

Industry analysts see this as a natural evolution of GitHub’s “AI pair programmer” vision. “The original Copilot inserted code snippets; the new CLI agent handles entire workflows,” says Forrester analyst David Truog. “That blurs the line between IDE and terminal, and challenges Microsoft to keep the experience consistent across VS Code, GitHub, and the command line.”

Getting Started Today

To get the new Copilot CLI on your Windows machine:

  1. Update to the latest GitHub CLI: winget install GitHub.cli (or choco upgrade gh).
  2. Authenticate: gh auth login and follow the prompts.
  3. Install the Copilot extension: gh extension install github/gh-copilot.
  4. Initiate the agent workbench: gh copilot init-agent.
  5. Launch the interface: gh copilot.

The first launch will offer a brief guided tour that highlights key commands: Ctrl+T to switch tabs, Ctrl+K to open the command palette, and Ctrl+Enter to submit a prompt to the agent. Users can customize keybindings by editing %APPDATA%\\gh\\config.yml.

For teams evaluating enterprise rollout, GitHub has published a Windows administration guide covering group policy objects for Copilot settings, proxy configuration, and logging. The guide also details how to deploy the agent workbench via Microsoft Intune, ensuring compliance with corporate security policies.

With this release, GitHub has transformed the humble CLI from a mere command executor into a collaborative, intelligent workspace. Windows developers now have a powerful new tool that can handle the drudgery of project management, code review, and environment setup—freeing them to focus on creative problem-solving. The agent workbench, in particular, hints at a future where the terminal is no longer just a tool you use, but a partner you collaborate with.