GitHub Copilot
An AI pair programmer by GitHub and OpenAI that suggests code completions, generates functions from comments, and — in agent mode — autonomously implements features across a codebase.
GitHub Copilot, launched in 2021 and built on OpenAI Codex (later GPT-4), was the first AI pair programmer to reach mainstream adoption. It integrates as an extension into VS Code, JetBrains, Neovim, and Visual Studio, making it the broadest-reaching AI coding tool by editor support.
Core Capabilities
- Inline completions — suggests the next line or block as you type, shown as ghost text
- Copilot Chat — context-aware conversation about your code (explain, refactor, generate tests)
- Slash commands —
/explain,/fix,/tests,/docfor common tasks - Workspace context — Copilot Chat can reference your entire open workspace, not just the current file
- Copilot Edits — propose changes across multiple files from a single prompt
- Copilot Agent (Coding Agent) — agentic mode where Copilot autonomously implements GitHub Issues
How It Works
Copilot sends your current file, surrounding context, and any chat messages to a hosted LLM (model choice varies by plan — GPT-4o, Claude, Gemini). The response is streamed back and displayed as suggestions. For Copilot Chat, it uses retrieval over your workspace files to find relevant context before calling the model.
Copilot vs Cursor
Cursor offers deeper agentic capabilities and a more opinionated editor experience built from scratch around AI. Copilot's advantage is ecosystem: it works wherever developers already are. For a team that can't switch editors but wants AI assistance, Copilot is the practical choice. For a team willing to adopt a new editor for maximum AI leverage, Cursor or Windsurf offer more.
Context Window and Privacy
Copilot sends code context to GitHub's servers. For organizations with strict data requirements, Copilot for Business offers options to disable model training on your code. Alternatively, self-hosted options like Ollama or Vllm behind a code-completion proxy keep code on-premises.
Prompt Engineering for Copilot
Copilot works best when you prime it with context: comments explaining intent, type annotations, and meaningful function/variable names. A well-named function with a docstring will get better completions than an ambiguous one — this is Prompt Engineering applied at the code-as-prompt level.
Related Terms
- Cursor — AI-first editor with stronger agent capabilities
- Windsurf — alternative AI IDE
- Prompt Engineering — the skill that makes Copilot suggestions more accurate
- Llm — the underlying technology powering suggestions