GitHub Copilot CLI

AI-powered coding assistant in the terminal. Uses the same agentic engine as GitHub's Copilot coding agent.

Official docs

Install

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

# macOS & Linux
$ curl -fsSL https://gh.io/copilot-install | bash

# Homebrew
$ brew install copilot-cli

# npm
$ npm install -g @github/copilot

# Windows (WinGet)
$ winget install GitHub.Copilot

Install a specific version to a custom path:

1

$ curl -fsSL https://gh.io/copilot-install | VERSION="v0.0.369" PREFIX="$HOME/custom" bash

Launch & authenticate

1

$ copilot

On first launch, use /login to authenticate with GitHub. You need an active Copilot subscription.

You can also authenticate via a fine-grained PAT with the "Copilot Requests" permission, exported as GH_TOKEN or GITHUB_TOKEN.

Keyboard shortcuts

Slash commands

Session

Code & GitHub

Config

Plan mode

Plan mode lets you structure an implementation before writing any code. Models achieve higher success rates when following a concrete plan.

Activate plan mode

• Shift+Tab to toggle between normal mode and plan mode
• Or use /plan directly from normal mode:

1

/plan Add OAuth2 authentication with Google and GitHub providers

What happens

1. Copilot analyzes the request and the codebase
2. Asks clarifying questions to align on requirements
3. Creates a structured plan with checkboxes in plan.md
4. Waits for your approval before implementing

You can edit the plan with Ctrl+Y before starting implementation.

Recommended workflow

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14

# 1. Explore the existing code (read-only)
Read the authentication files but don't write code yet

# 2. Plan
/plan Implement password reset flow

# 3. Review and adjust the plan, then implement
Proceed with the plan

# 4. Verify
Run the tests and fix any failures

# 5. Commit
Commit these changes with a descriptive message

When to use /plan

Fleet mode

Fleet mode lets Copilot break a complex task into subtasks executed in parallel by subagents. Each subagent has its own context window.

Usage

Prefix your prompt with /fleet:

1

/fleet Refactor all API endpoints to use the new middleware, add tests for each

How it works

1. The main agent analyzes the prompt and identifies independent subtasks
2. It orchestrates subagents, each working on its part in parallel
3. Results are gathered once all subagents are done

When to use /fleet

Combining with plan + autopilot

The most powerful workflow for large tasks:

1. Shift+Tab → plan mode, create the implementation plan
2. Notice the plan has parallelizable elements
3. Select Accept plan and build on autopilot + /fleet

Choose model per subtask

By default subagents use a lightweight model. You can specify a model in the prompt:

1
2
3

/fleet Use Claude Opus 4.5 to analyze the auth module.
Use GPT-5.3-Codex to generate the migration scripts.
Use @test-writer to create unit tests.

Custom instructions

Copilot reads instructions from these files (in order):

• CLAUDE.md, GEMINI.md, AGENTS.md (in git root & cwd)
• .github/instructions/**/*.instructions.md
• .github/copilot-instructions.md
• $HOME/.copilot/copilot-instructions.md

MCP servers

Copilot ships with GitHub's MCP server by default. Add custom MCP servers via /mcp or by editing the config directly.

LSP support

Copilot CLI supports Language Server Protocol for code intelligence (go-to-definition, hover, diagnostics). Servers are not bundled — install them separately.

Config at user level (~/.copilot/lsp-config.json) or repo level (.github/lsp.json):

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

{
  "lspServers": {
    "typescript": {
      "command": "typescript-language-server",
      "args": ["--stdio"],
      "fileExtensions": {
        ".ts": "typescript",
        ".tsx": "typescript"
      }
    }
  }
}

Update

1
2
3

$ copilot
# then
/update