GitHub Copilot review
★ 4.2/5AI coding assistant
AI coding agent and pair programmer embedded in your IDE, terminal, and GitHub. Offers inline code completions, chat, an autonomous agent mode, a cloud-based coding agent that opens pull requests, and automated code review, with a choice of models from OpenAI, Anthropic, and Google.
Affiliate link · how we make money
Best for in-IDE coding. GitHub Copilot starts $10/mo (Pro) on a per-seat + usage credits model. Pro at $10/mo includes $10 of AI Credits = 1,000 credits (1 credit = $0.01). Code completions are free and unlimited, so credits only burn on agent mode, chat, code review, and the cloud agent, billed by token usage at each model's API rate. A cheap model like GPT-5 mini ($0.25/1M input tokens) lets a single chat costing ~10k input + 2k output tokens consume well under 1 credit; a Claude Opus 4.8
Spec
The gotcha
Billing changed on June 1, 2026: flat "premium request" counting is gone, replaced by token-metered AI Credits priced at each model's real API rate, so heavy use of expensive models (Claude Opus, GPT-5) can burn through your included pool far faster than the old fixed-request model suggested, and the old cheaper-model "fallback" when you ran out is being removed, so once credits are spent work simply stops unless you allow overage at published rates. Code completions stay free, but chat, agent mode, and code review all meter tokens (including cached and output tokens), and Copilot code review also eats GitHub Actions minutes. Also, since April 20, 2026, new sign-ups for the Pro, Pro+, and Student individual plans were paused, and Opus models were pulled from Pro and now require Pro+ or higher.
Decode the real cost →Best for
- You want an AI coding agent that lives directly inside VS Code, Visual Studio, JetBrains, the CLI, and GitHub itself, completing code as you type, running multi-step tasks autonomously in agent mode, and delegating whole pull requests to a cloud agent that codes and reviews on your behalf.
source: github.com/pricing →
GitHub Copilot pricing tiers
Prices move. Cells flagged verify link to the live vendor page.
| tier | monthly | annual | included | unit |
|---|---|---|---|---|
| Free | $0 | $0 | 2,000 code completions + 50 agent/chat requests per month; access to Haiku 4.5, GPT-5 mini and other models; Copilot CLI | per user / month |
| Pro | $10verify | $100 | Everything in Free plus unlimited completions, code review, Copilot cloud agent; ~$10 in monthly AI Credits (legacy: 300 premium requests, ~1,000 base + 500 flex credits) | per user / month |
| Pro+ | $39verify | $390 | Everything in Pro plus all models incl. Claude Opus, GitHub Spark, third-party agent delegation; ~$39 in monthly AI Credits (~7,000 incl. 3,900 base + 3,100 flex; legacy 1,500 premium requests) | per user / month |
| Max | $100verify | n/a | Highest individual allowance: ~$100 base + ~$100 flex AI Credits (~20,000 credits total); priority access to newest models and features | per user / month |
| Business | $19verify | n/a | Org management, policy controls, pooled credits; ~$19 in monthly AI Credits (promo boost to ~$30 Jun-Aug 2026) | per user / month |
// pools and per-unit rates are volatile · cells flagged verify link to the live vendor page
Compare GitHub Copilot with…
Questions
How much does GitHub Copilot cost?
GitHub Copilot starts from $10/mo (Pro). Pro at $10/mo includes $10 of AI Credits = 1,000 credits (1 credit = $0.01). Code completions are free and unlimited, so credits only burn on agent mode, chat, code review, and the cloud agent, billed by token usage at each model's API rate. A cheap model like GPT-5 mini ($0.25/1M input tokens) lets a single chat costing ~10k input + 2k output tokens consume well under 1 credit; a Claude Opus 4.8
Can you self-host GitHub Copilot?
No, GitHub Copilot is cloud-only.
Does GitHub Copilot support MCP?
GitHub Copilot supports MCP as both client and server, so it works with assistants like Claude and ChatGPT.
What's the catch with GitHub Copilot's pricing?
Billing changed on June 1, 2026: flat "premium request" counting is gone, replaced by token-metered AI Credits priced at each model's real API rate, so heavy use of expensive models (Claude Opus, GPT-5) can burn through your included pool far faster than the old fixed-request model suggested, and the old cheaper-model "fallback" when you ran out is being removed, so once credits are spent work simply stops unless you allow overage at published rates. Code completions stay free, but chat, agent mode, and code review all meter tokens (including cached and output tokens), and Copilot code review also eats GitHub Actions minutes. Also, since April 20, 2026, new sign-ups for the Pro, Pro+, and Student individual plans were paused, and Opus models were pulled from Pro and now require Pro+ or higher.