GitHub Copilot CLI has reached General Availability (GA) and is now available to all paid Copilot users.
If you last looked at this tool when it was still just a question-answering assistant in the command line, it has evolved significantly since then.
What the CLI Is Now
It is no longer just a helper that answers questions in the terminal.
Recent versions have added substantial new capabilities:
v1.0.21 (April 7): Added the mcp command, allowing users to manage MCP servers directly from the CLI — connect, view, and switch between them. MCP is now essentially the universal interface standard for AI tools, and integrating it into Copilot CLI means the tool can plug into any existing MCP ecosystem.
v1.0.18 (April 4): Introduced the Critic agent — an automatic agent that reviews your plan or complex task before execution, identifying potential issues. It functions like an automated code reviewer, but operates before execution begins.
Dual-mode operation:
- Plan mode: Analyzes requirements, clarifies questions, generates an execution plan, and waits for your confirmation before proceeding. Suitable for tasks with high uncertainty.
- Autopilot mode: Runs fully automatically without pausing for confirmation, capable of executing commands, iterating on code, and self-correcting errors.
Model selection: Claude Opus 4.6, Claude Sonnet 4.6, GPT-5.3-Codex, Gemini 3 Pro, Claude Haiku 4.5 — choose based on task complexity and budget.
Copilot SDK Also Arrives
On April 2, GitHub quietly released the Copilot SDK (public preview).
This may have greater long-term significance than the CLI GA: The SDK exposes the same runtime that powers Copilot CLI and the cloud agent, allowing developers to embed this agent into their own applications.
This is the first time GitHub has opened up Copilot's agent capabilities as embeddable infrastructure.
What This Direction Means
Over the past two to three years, GitHub Copilot has evolved from autocomplete tab completions to handling issues, reviewing code, and now to a full-featured CLI agent.
Each step expands the control plane — from IDE to browser extension, to PR review, to the command line, and now to the SDK.
The current form: Give Copilot CLI a task, and it can plan, execute, check its own results, and connect to MCP tools — all without leaving the terminal.
For developers who prefer keyboard-centric workflows, this trajectory is highly appealing. GitHub Copilot now has over 20 million users, and a significant portion of that base are heavy terminal users.
Copilot CLI's direct competitors are Claude Code and Gemini CLI. All three are pushing in this direction, and whichever gets developers to form a habit will win this entry point.
Sources: GitHub Copilot CLI Reaches General Availability, CocoLoop; Bringing Agentic Coding to the Terminal (Visual Studio Magazine); RELEASE: GitHub Copilot CLI – AI for the Command Line (kurtsh.com)