GitHub Copilot CLI vs Claude Code Comparison

# GitHub Copilot CLI — install and basic usage (official docs)
npm install -g @github/copilot

# Start an interactive session (authenticates with your current GitHub account)
copilot

# Run with a one-off prompt
copilot -p "Find and fix the failing tests in this repo"

# Select a specific model (switch across providers)
copilot --model claude-sonnet-5 -p "Update the README"

# Enable the local sandbox from within a session: /sandbox enable  ·  for a cloud session: copilot --cloud

# Add an MCP server (remote HTTP server)
copilot mcp add --transport http notion https://mcp.notion.com/mcp
# The GitHub MCP server ships built into Copilot CLI, no extra configuration needed

# Open a PR in natural language
copilot -p "Open a pull request for these changes"

# Claude Code — install and basic usage (official docs)
npm install -g @anthropic-ai/claude-code

# Start an interactive session in the project directory
claude

# Run with a one-off (headless) prompt
claude -p "Find and fix the failing tests in this package"

# Select a model (within the Anthropic family)
claude --model claude-opus-5

# .mcp.json — project-level MCP server definition
# The GitHub remote MCP server expects a PAT in the Authorization header for auth
{
  "mcpServers": {
    "github": {
      "type": "http",
      "url": "https://api.githubcopilot.com/mcp/",
      "headers": { "Authorization": "Bearer YOUR_GITHUB_PAT" }
    }
  }
}

# settings.json — sandbox and permission policy
{
  "sandbox": { "enabled": true },
  "permissions": { "allow": ["Bash(npm run test:*)"] }
}