GitHub Copilot
GitHub Copilot started as autocomplete for code, and GitHub now calls it an AI coding agent. You can still use it for inline suggestions as you type, but most of the newer features are about handing Copilot a whole task and reviewing what it produces before anything is merged.
Copilot runs in VS Code, Visual Studio, JetBrains IDEs, Xcode, Neovim, Eclipse and Zed, in the terminal through Copilot CLI, in a desktop app, and on github.com and GitHub Mobile. Because it's built into GitHub, it can work straight from an issue or a pull request with the repository context already loaded.

Copilot code review suggesting a fix for a race condition in a pull request. Image from GitHub.
Code completion and agent mode
Inline completions and next edit suggestions are the part of Copilot most developers know. They're unlimited on every paid plan, and the free plan includes 2,000 completions a month. Agent mode in the editor goes further: you describe a change, and Copilot edits files across the project and runs commands, checking its own work as it goes.
Delegating tasks and running sessions
The bigger shift is delegation. You can assign an issue to Copilot, or start a session from an issue or pull request, and it plans and writes the code in the cloud, then opens a pull request for you to review. Each task gets its own session, so several can run at once, and you can check on a long-running one from the desktop app, the CLI, your editor or your phone.
Automations run reusable agent tasks on a schedule or from a trigger, such as triaging new issues every morning. On Pro+ and Max you can also hand tasks to third-party coding agents, including Claude from Anthropic and OpenAI Codex, from inside Copilot.
Code review
Copilot code review comments on pull requests in GitHub and on file diffs in your editor, with suggested changes you can commit directly. It can also review local changes for code quality and security problems before you open a pull request, then fix the findings you pick. Code review on GitHub uses GitHub Actions minutes as well as Copilot credits.
Models and customization
You can let Copilot pick a model for each task or choose one yourself from a catalog of more than 30 models. Which models you get depends on your plan, and the free plan is limited to lighter models. You can also connect your own API key for a supported cloud provider, or run models locally with Ollama or Foundry Local.
Repository instructions and skills set coding standards for Copilot to follow, and MCP servers and Copilot Connectors bring in tools and data from other services.
Plans and pricing
Copilot bills AI usage in GitHub AI Credits. Chat, agent mode, code review, the cloud agent and Copilot CLI all use credits, and the amount depends on the model and the size of the task. Code completions don't use credits. When you run out, you can wait for the monthly reset or set a budget for extra usage. Switching to a lighter model also makes credits last longer.
| Plan | What you get |
|---|---|
| Free | 2,000 completions a month, lighter models, limited chat and agent usage, and Copilot CLI |
| Pro | Unlimited completions, the cloud agent and code review, third-party agents, model selection and a monthly credit allowance |
| Pro+ | Everything in Pro, plus premium models including Claude Opus, audit logs and about four times Pro's included usage |
| Max | Priority access to new models and features, and about three times Pro+'s included usage, for heavy agent work |
| Business | Credits pooled across the organization, access controls, budget controls, IP indemnity and data privacy terms |
| Enterprise | Everything in Business, with more credits per user and priority access to new models |
Verified students can get a free Copilot Student plan, and GitHub Enterprise now includes access to Copilot.
Is GitHub Copilot free?
Yes, there's a free plan with no credit card required. It includes 2,000 code completions a month, entry-level models, Copilot CLI and limited chat and agent use. It's enough to try Copilot properly, but regular agent work needs a paid plan.
Does GitHub train on my code?
It depends on your plan. On Free, Pro and Pro+, GitHub may use your Copilot interaction data, such as prompts and generated code, to train its models, but you can turn this off at any time in your Copilot settings without losing any features. Business and Enterprise data isn't used that way.
GitHub Copilot vs Cursor?
Cursor is a separate AI code editor built on VS Code, so you switch your whole editor to use it. Copilot works inside the editors you already have and connects directly to GitHub issues, pull requests and code review, which suits teams whose work already runs through GitHub.
Alternatives to GitHub Copilot
- Cursor: VS Code-based AI code editor with tab completion and an agent that understands your whole codebase.
- OpenAI Codex: Writes, tests and reviews code from the ChatGPT desktop app, the terminal, your IDE and the cloud.
- Windsurf: AI-native IDE with Cascade, an agent that edits multiple files and runs terminal commands.
- Cline: Free, open-source coding agent for VS Code and JetBrains that uses your own API key and asks before every action.
- Jules: Google's asynchronous coding agent that works on tasks in your GitHub repositories in the background.

TRY IT