GitHub Copilot Review

What GitHub Copilot actually does

Copilot is the original AI coding assistant — autocomplete, chat, and (more recently) agent-style multi-step task execution — integrated into the major editors and JetBrains IDEs. The product has expanded substantially from its original GPT-3.5-powered autocomplete origins. Current Copilot includes inline completions, a sidebar chat with codebase awareness, slash commands, pull request review summaries, and an agent mode that can edit multiple files and run terminal commands with permission. Copilot is no longer a single-model product — users can now choose the underlying model for chat and agent tasks, including GPT-4o, Claude, and Gemini variants, depending on plan.

The strategic positioning has shifted. Copilot used to be the obvious choice because there were no real competitors. With Cursor and Windsurf in the market, Copilot competes on enterprise governance, GitHub native integration, and broad editor support rather than on raw capability per dollar.

What works well

Editor coverage is the practical advantage. Copilot is supported across VS Code, all JetBrains IDEs, Visual Studio, Neovim, Eclipse, and Xcode. For teams with mixed toolchains, this matters — Cursor's VS Code-only constraint is a real limitation.

Enterprise governance has matured. The Copilot Enterprise tier offers private model selection, code reference filtering (to reduce the chance of generating verbatim copies of training data), audit logging, and SSO. For regulated industries, this is the only AI coding tool that has cleared most procurement processes.

Pull request summaries and reviews are an underrated feature. Copilot can summarize what a PR changed, suggest review comments, and generate test cases for the diff. For teams with thin code review capacity, this raises the floor on what gets reviewed at all.

Where it falls short

The autocomplete and chat experiences trail Cursor in raw quality and latency. The gap is small but consistent in side-by-side comparisons. Engineers who switch from Cursor to Copilot generally notice the regression within a week.

Multi-file refactoring is weaker than Cursor's Composer. The agent mode improves this, but it's still in catch-up territory rather than ahead.

Pricing for individuals is competitive ($10/mo) but Business and Enterprise tiers are noticeably more expensive than Cursor's equivalents. The premium pays for governance features that smaller teams don't need. Additionally, all tiers now come with a monthly allowance of "premium requests" (used for chat, agent tasks, and premium model access); usage beyond the allowance is billed per-request unless the org enables metered overage. Base code completions remain unlimited, but heavy agent-mode users should expect to watch their premium request quota.

Who should use it

Copilot is the right pick for engineers who can't or won't move to Cursor — JetBrains users, Visual Studio users, anyone whose company has standardized on it. It's also the right pick for enterprise teams where procurement has already approved Copilot but not Cursor.

Engineers using VS Code as their primary editor, with no enterprise constraints, will likely get more value from Cursor at the same or lower price point.

Pricing notes

Individual ($10/mo or $100/year, 300 premium requests/mo), Business ($19/seat/mo, 300 premium requests/mo), Enterprise ($39/seat/mo, 1,000 premium requests/mo). All tiers include unlimited standard code completions; premium requests cover chat, agent mode, and non-default model usage, with optional metered billing for overages. The Business tier is the right starting point for small teams; Enterprise adds the governance and customization features.

The architecture-decision gap

Copilot helps you write code faster. It does not decide what to build, what to deprecate, what to extract into a service, or what to leave alone. Those system-level decisions still require senior engineers or an architecture-aware agent layer that reads your codebase as a whole and proposes changes at the program level.