Overview
What GitHub Copilot is and what it can do in VS Code
Read time: 4 min
title: "Overview" description: "What GitHub Copilot is and what it can do in VS Code" section: "Copilot" readTime: "4 min"
GitHub Copilot Overview
GitHub Copilot brings AI agents to Visual Studio Code. Describe what you want to build, and an agent plans the approach, writes the code, and verifies the result across your entire project.
What Copilot Can Do
- Build features end-to-end — describe a feature in natural language; the agent scaffolds, implements across files, and runs tests
- Debug and fix failing tests — agent reads the error, traces the root cause, applies a fix, re-runs tests
- Refactor and migrate codebases — coordinated changes across files with build verification
- Test web apps (Experimental) — open app in integrated browser, verify features, take screenshots
- Create PRs for team review — delegate to a cloud agent that branches, implements, and opens a pull request
Three Core AI Surfaces
| Surface | How to access | Best for |
|---|---|---|
| Agents | Ctrl+Shift+I in Chat view | Multi-file, autonomous tasks |
| Inline suggestions | Type in editor (ghost text) | Completions as you code |
| Inline chat | Ctrl+I in editor | Targeted edits without context switch |
| Smart actions | Right-click menu, Source Control sparkle | Commit messages, fix errors, generate docs |
Agents Overview
An agent takes a goal, breaks it into steps, edits files across your project, runs commands, and self-corrects when something goes wrong. Each task runs inside an agent session you can track, pause, resume, or hand off.
Agent types:
- Local — interactive, full workspace access
- Background (Copilot CLI) — autonomous, runs while you keep working
- Cloud — remote, integrates with GitHub pull requests
- Third-party — Anthropic Claude, OpenAI Codex
Built-in agents:
- Agent — autonomously plans and implements
- Plan — creates structured implementation plan before coding
- Ask — answers questions without making file changes
Customization Options
| Option | Purpose |
|---|---|
| Custom instructions | Project-wide coding standards |
| Prompt files | Reusable prompts for common tasks |
| Custom agents | Specialized personas with specific tools |
| Agent skills | Domain-specific capabilities |
| MCP servers | Connect to external APIs, databases |
| Hooks | Run commands at lifecycle events |
Pricing
- Free plan — monthly limit on inline suggestions and chat interactions
- Paid plans — more interactions, premium models, remote indexing, code review
Run /init in chat to generate custom instructions for your project automatically.