GitHub Copilot is the AI coding assistant baked into VS Code, Visual Studio, JetBrains IDEs, Xcode, Neovim, the GitHub CLI, GitHub Mobile, and now its own standalone desktop app. It writes inline suggestions as you type, answers questions about your codebase in chat, and can run as a semi-autonomous agent that opens pull requests on its own. As of June 1, 2026, GitHub tore up its old "premium request" billing and replaced it with a straight credit-pool system — every plan now comes with a dollar-denominated bucket of AI Credits that drains based on actual token usage. If you're trying to figure out whether the $10, $39, or $100 tier still makes sense, the math changed and it's worth walking through.
Key Features
- Inline code completions and Next Edit Suggestions are unlimited on every paid plan and cost zero credits — the everyday autocomplete-as-you-type experience isn't affected by the new billing at all.
- Agent mode can take a GitHub issue, write the code, run tests, and open a pull request with no hand-holding — but it's also the single fastest way to burn through a credit pool, since one agent session can use roughly a thousand times the tokens of a simple chat question.
- Model choice is wide open: Claude, GPT, and Gemini models are all available from the same interface, and each one draws credits at a different rate depending on its published per-token API cost.
- A new desktop app (macOS, Windows, and Linux) runs agent sessions in isolated workspaces outside your editor, so you can kick off a background task, keep coding elsewhere, and come back to review the diff.
Screenshots
Why You'll Love GitHub Copilot
The credit math is more transparent than premium requests ever were
The old system charged a flat "1 request" no matter how long or short your prompt was, then applied a confusing multiplier depending on which model you picked. AI Credits fix the opacity problem: 1 credit equals exactly $0.01, and you can watch your balance drop in real time as tokens get consumed. Pro's $10/month plan includes $10 in base credits plus a $5 "flex" allotment, for $15 total. Pro+ at $39 gets $39 base plus $31 flex, for $70 total. Max at $100 gets $100 base plus a full $100 flex match, for $200 total. The flex portion is explicitly variable — GitHub says it will shift as model pricing and efficiency change, so don't treat this month's total as a permanent number.
Agent mode is powerful, but it's the thing that empties your wallet
Every credit horror story circulating since June has the same root cause: agent-driven workflows. A single autonomous session — the kind that reads a whole repo, plans changes, writes code, and iterates on test failures — can chew through $6 to $12 or more in one go, because it's making dozens of model calls with large context windows instead of one short chat reply. If you mostly want autocomplete and the occasional chat question, $15–$70 in monthly credits goes a long way. If you're running agent mode daily against a large codebase, even the Max plan's $200 can disappear inside a couple of weeks. There's no default spending cap either — GitHub added a manual spending-limit toggle in Settings → Billing → GitHub Copilot in July 2026, and setting it to $0 is the only way to guarantee you won't get billed for overage once your included credits run out.
Category: AI Coding Assistant
Platform: Windows, Mac, Linux, iOS, Android, Web (VS Code, Visual Studio, JetBrains, Xcode, Neovim, CLI)
Price: Free tier available; Pro $10/mo, Pro+ $39/mo, Max $100/mo, Business $19/user/mo
Frequently Asked Questions
Yes, there's a genuine free tier for individual developers: 2,000 code completions and 50 chat requests a month, plus limited agent access. Once you hit those caps, code completions keep working in a reduced form but chat and agent features stop until the next monthly reset, or you upgrade to Pro at $10/month.
No. Every feature, from inline completions to chat to agent mode, sends requests to GitHub's servers to run the underlying model, so you need an active internet connection the entire time you're using it.
In an editor, Copilot needs read access to the files and repository you're working in so it can generate relevant suggestions. Agent mode goes further, requesting permission to write files, run terminal commands, and open pull requests on your behalf — each of those actions is scoped to the repos you explicitly grant it access to, and you can review every change before it merges.