GitHub Copilot

GitHub Copilot is GitHub’s AI coding assistant. It suggests code as you type, answers questions in chat, edits across several files, helps review pull requests and, on supported plans, runs agents that work on tasks such as GitHub issues. It works inside popular code editors and on GitHub itself.

Quick facts

• Made by: GitHub, which is owned by Microsoft
• Free plan: Yes, with monthly limits on code completions and premium requests
• Paid plans: Pro, Pro+ and Max for individuals, plus Business and Enterprise for organisations
• Works in: VS Code, Visual Studio, JetBrains IDEs and Neovim, plus GitHub.com, GitHub Mobile and the command line
• Pricing last checked: October 2026, using several recent published sources. Plan allowances and credit rules changed on 1 June 2026, so confirm on GitHub’s official plans page.

What GitHub Copilot is best at

• Autocomplete inside your editor: inline suggestions for lines, functions and boilerplate while you type.
• Chat about your code: ask what a function does, why a test fails or how to refactor a file.
• Agent tasks: hand a well-described issue to an agent and review the changes it proposes.
• Working inside GitHub: pull request summaries, code review help and repository-aware answers for teams already on GitHub.
• Choice of models: several providers’ models are available, with the more powerful ones drawing from your premium allowance.

Plans and pricing

Base prices stayed the same when GitHub moved to usage-based billing on 1 June 2026. Advanced use, such as agents and premium models, is now metered as credits on top of the base fee, so heavy agent users can pay more than the flat price. Published sources disagree on some allowance numbers, so treat the figures above as a guide and confirm on GitHub’s official page. GitHub has also offered free access to verified students, teachers and some open-source maintainers, so check whether you qualify.

Best for

• Developers who want AI help without leaving their usual editor.
• Teams already on GitHub who want one vendor for code hosting and AI assistance.
• Organisations that need audit logs, policy controls and a single bill.

Less suited to

• People who want fixed, predictable monthly costs for heavy agent use, since credits are metered.
• Developers who prefer a whole editor built around AI. Cursor takes that approach.
• Anyone who needs a no-code way to build an app from a plain-language description, where Lovable is closer.

How to get started

1. Sign in to GitHub and open the Copilot page to start on the Free plan.
2. Install the Copilot extension in your editor, such as VS Code or a JetBrains IDE, and sign in.
3. Try it on a real task: ask Copilot to explain a function, write tests for it or suggest a refactor.
4. Review every suggestion before you accept it, and run your tests.
5. If you hit the Free limits, compare Pro against your actual usage before you upgrade. Organisations should check Business policy controls first.

Prompt ideas to copy

Explain what the selected function does in plain English, then list any edge cases or inputs it does not handle.

Write unit tests for the selected function using [test framework]. Cover normal cases, edge cases and error handling, and keep each test small.

Refactor this file to reduce duplication without changing behaviour. Add a short comment for each change explaining why.

Fix issue #[number]: [describe the problem]. Make the smallest change that works, add a test that fails without it, and summarise what you changed.

Limitations and things to watch

• Suggestions can be wrong, outdated or insecure, so review and test everything before it ships.
• Credits and allowances changed in June 2026, and heavy agent use can exceed what your plan includes.
• Check your organisation’s rules before pasting private or licensed code into any AI tool.
• Plan names, prices and allowances change often, so confirm before you commit.

Alternatives

Compare Cursor for an AI-first editor, Claude and ChatGPT for general coding help in chat, and Lovable for building web apps from prompts.