GitHub Copilot

What it is

GitHub Copilot is GitHub’s own AI pair programmer: inline code completions, a chat panel, an agent mode that edits across files and runs terminal commands, a cloud agent for longer-running tasks, a CLI, and an automated code review step, all wired into the same account as a developer’s repositories. It ships as an extension for VS Code, Visual Studio, JetBrains IDEs, Vim/Neovim and Xcode, plus a standalone CLI and a chat surface on github.com itself.

The product’s packaging changed repeatedly in 2026. A June 1 billing change replaced the old “premium requests per month” quota with a credit system, where $0.01 equals one credit and completions stay free while chat, agent mode, cloud agent and code review draw the balance down. Weeks earlier, on April 20, GitHub had paused new individual sign-ups for Pro, Pro+ and Student and pulled Opus models out of Pro, citing infrastructure strain as more users ran agents rather than just accepting completions. Sign-ups and a new $100/mo Max tier returned in stages between June 17 and June 29.

Who should use it

A developer who already works inside GitHub and wants AI assistance to show up in the same editor and account rather than a separate subscription. It is a strong fit for completions-first use, since those are unlimited on every paid plan regardless of how much chat or agent usage a credit balance allows. It fits less well for anyone whose work is mostly large, multi-file agentic refactors, where independent reviews consistently rank Copilot’s agent behind Cursor and Claude Code.

What stands out

The credit math is the thing to actually understand before picking a tier. Pro’s $10/mo buys $15 of credits ($10 base, $5 “flex”), Pro+’s $39/mo buys $70, and Max’s $100/mo buys $200, so every paid tier includes more usage value than its sticker price. None of that credit balance touches code completions or next-edit suggestions, which GitHub keeps free and unlimited across Free, Pro, Pro+ and Max alike; it only meters chat, agent mode, the cloud agent, and code review, each pulling from the balance at a rate that depends on which model answers the request.

Which model you can reach has also moved during the year. Opus models left the Pro tier in the April pause and stayed exclusive to Pro+ when sign-ups reopened, while Free, Pro, Pro+ and Max all otherwise draw from the same roster: Anthropic Claude, OpenAI GPT, Google Gemini, xAI Grok, Microsoft MAI and Kimi. Existing Pro and Pro+ subscribers who didn’t want the April changes had a window through May 20, 2026 to cancel for a prorated refund, a sign of how much changed for people who had already paid for the plan.

Team pricing sits at $19/user/month for Business and $39/user/month for Enterprise, both pooling credits across the organization rather than per seat. GitHub’s own changelog still lists self-serve Business sign-ups as paused for organizations on the free GitHub or Team plans since April 22, 2026, with no reopening confirmed as of this writing, which is worth checking directly before assuming a team account is purchasable today.

The verdict

For a developer who lives in GitHub already, Copilot removes the friction of adding another tool: completions are free on every tier, Pro is the cheapest seat among the big AI coding products at $10/mo, and the credit balances on paper buy more than their cash price. What it asks in return is tolerance for a product whose billing and model access keep moving. In one year it switched from premium requests to credits, paused and reopened individual sign-ups, pulled a flagship model out of its cheapest paid tier, and is still showing a paused self-serve status for some team sign-ups.

Skip Copilot specifically for heavy, large-scope agentic work: the reviews that compare agents head to head put Cursor and Claude Code ahead of it there, and both are already covered in this directory. Pick Copilot instead for the everyday case, completions plus chat plus occasional agent use inside a GitHub-centric workflow, and budget for the credit system rather than assuming last year’s premium-request numbers still apply.