GitHub Copilot is an AI-powered coding assistant made by GitHub (owned by Microsoft) that sits inside your code editor and helps you write software faster. It watches what you type, understands the context of your project, and suggests lines or entire blocks of code in real time. Since its launch in 2021 — originally built on OpenAI’s Codex model — it has grown from simple autocomplete into a full-featured platform that can review pull requests, answer questions in a chat window, and autonomously plan and execute multi-step coding tasks.
How GitHub Copilot Works
Copilot works by analyzing the code around your cursor — the lines above and below, the file you’re editing, and context from other open files — then sending that context to an AI model in the cloud. The model predicts what code is most likely to come next and returns a suggestion instantly.
Unlike a search engine or code-snippet library, Copilot does not copy code from GitHub repositories. Its models were trained on public source code and natural language, but the suggestions it generates are new — probabilistic completions based on patterns learned during training. According to GitHub, only a small fraction of outputs closely resemble training data.
Today Copilot supports a menu of AI models, including OpenAI’s GPT series, Anthropic’s Claude, Google’s Gemini, and xAI’s Grok. Developers on paid plans can choose which model powers their session.
What GitHub Copilot Can Do
Code completion. As you type, Copilot suggests the next line, the rest of a function, or an entire class. Pressing Tab accepts the suggestion; pressing Escape dismisses it.
Copilot Chat. A conversational window lets you ask questions — “explain this function,” “write a unit test for this method,” “what’s wrong with this loop?” — and get answers grounded in your current file or project.
Code review. On GitHub.com, Copilot can automatically review pull requests, flagging bugs, summarizing changes, and generating PR descriptions.
Agent mode. Introduced in 2025, agent mode lets Copilot work autonomously: you describe a task in plain language, and the assistant plans the steps, edits files, runs commands, and iterates until the job is done. This is what makes agentic billing matter — complex autonomous tasks can consume thousands of tokens per run, far more than simple suggestions.
Next Edit Suggestions. Rather than only completing the code at your cursor, Copilot tracks your recent edits and anticipates the next change you’ll likely want to make elsewhere in the file.
Which IDEs and Languages It Supports
Copilot integrates with Visual Studio Code (most features), JetBrains IDEs (IntelliJ, PyCharm, WebStorm, and others), Visual Studio, Vim/Neovim, Eclipse, and the GitHub CLI. A mobile app and the GitHub website offer Copilot Chat.
It supports over 50 programming languages. Suggestion quality is strongest for languages best represented in public code: Python, JavaScript, TypeScript, Java, C#, Go, Ruby, and PHP.
How Much It Costs
GitHub Copilot switched to usage-based billing on 1 June 2026, replacing the older flat-rate subscriptions. The base plans include a monthly credit allowance; heavy agentic use can add charges beyond that allowance — which is what drove billing surprises for some developers this month.
As of June 2026, per GitHub’s official pricing page:
| Plan | Monthly price | What’s included |
|---|---|---|
| Free | $0 | 2,000 code completions, limited chat (Haiku 4.5 / GPT-5 mini) |
| Pro | $10 | Unlimited completions, $15 AI credits, premium model access |
| Pro+ | $39 | Unlimited completions, $70 AI credits, advanced models |
| Business | $19 per user | Pooled credits, governance controls, full model catalog |
| Enterprise | $39 per user | Priority features, 2× included credits vs. Business |
Code completions — the core autocomplete — remain unlimited on all paid plans and consume no credits. Credits are spent by Chat, agent-mode tasks, and code review, at $0.01 per credit with token rates varying by model.
In the News
When GitHub switched to usage-based billing in June 2026, some developers received invoices 10× to 50× larger than their previous flat-rate bills. The spike came from agentic tasks — long autonomous coding sessions that can burn thousands of tokens per run. GitHub clarified that agentic usage is opt-in and that the free tier is unaffected.
For the full story, see: GitHub Copilot’s First Usage-Based Bills Land.
FAQ
Is GitHub Copilot free?
Yes — there is a free tier that includes 2,000 code completions per month and limited chat access. Paid plans start at $10/month and add unlimited completions, more AI credits, and access to premium models.
Does GitHub Copilot work offline?
No. Copilot requires an internet connection; all suggestions are generated in the cloud by GitHub’s AI models.
Can GitHub Copilot write entire programs?
In agent mode it can plan and execute multi-step tasks — creating files, editing code, running tests — but it works best as a collaborator. You still review and approve the changes.
Is the code GitHub Copilot generates safe to use in any project?
Generally yes, but review AI-generated code before including it in projects with strict open-source licensing requirements. Ongoing litigation (Doe v. GitHub) is testing questions around training data and output in court, and the legal landscape continues to evolve.
Sources: GitHub Copilot documentation · GitHub Copilot pricing · Wikipedia: GitHub Copilot