GitHub Copilot

GitHub Copilot is an AI coding assistant built by GitHub (owned by Microsoft) and powered by models from Anthropic, OpenAI, Google, xAI, Moonshot AI and Microsoft, selected by the user. It operates as a plugin across VS Code, JetBrains IDEs, Neovim, Xcode, Eclipse, Visual Studio, Raycast, Zed and SQL Server Management Studio, and it also ships a CLI and a cloud agent. The combination of GitHub ecosystem depth and multi-IDE coverage is what separates it from AI-native editors such as Cursor, which ask developers to change editor. Copilot cloud agent creates pull requests from GitHub Issues and Copilot code review submits inline review comments. Model Context Protocol support runs in both directions. Copilot acts as an MCP client in agent mode on every plan including Free, with cloud agent MCP from Pro upward and allow-list governance on Business and Enterprise. GitHub also publishes its own official MCP server, open source under the MIT license, offered both as a remote server hosted by GitHub and as a local build. Its documentation names VS Code, Claude, Cursor, Windsurf, JetBrains and Xcode as supported clients. Third-party coding agents including Claude Code and OpenAI Codex can be delegated tasks from the CLI on every plan, and from the cloud agent on Pro+ and above. All prices are USD and the vendor publishes no annual rate, so every figure below is month to month. Copilot Free is $0 and includes 2,000 completions per month with limited chat and agent usage. A Copilot Student plan is offered free to eligible students. Pro is $10 per user per month. Pro+ is $39 and Max is $100, each adding premium model access and a larger included allowance. Business is $19 per user per month and Enterprise is $39. Both business cards carry a self-serve Get started path alongside Contact sales, and GitHub notes that new sign-ups are being enabled gradually. Billing uses GitHub AI Credits priced at $0.01 each: the included allowance is a base amount plus a flex allotment that GitHub states may change over time, giving $15 total on Pro, $70 on Pro+ and $200 on Max. Code completions and next edit suggestions are unlimited on all paid plans and are not billed in credits. The $10 Pro tier does not buy pooled organization credits, access control, organization-level budget governance, usage metrics, SAML SSO or IP indemnity, all of which start at Business. Audit logs start one tier lower, at Pro+. An individual on a paid plan can set a dollar budget for usage beyond the included allowance and Copilot then continues without interruption, with alerts at 75, 90 and 100 percent of that budget. With no budget set, usage stops when the allowance is spent. Restricting users to models carrying FedRAMP Moderate certification is a separate policy, available only on GitHub Enterprise Cloud with data residency in the US, disabled by default, and each request made under it consumes 110 AI credits. Data training also splits by plan. GitHub does not use Copilot Business or Enterprise data to train its models. The $10 Pro tier published here is an individual plan. Individual plans are used to train and improve AI models unless the subscriber opts out in Copilot settings, and GitHub states that opting out does not affect access to any Copilot feature. GitHub Copilot suits teams that want an assistant inside the editor they already use, with their source and workflow already on GitHub. Developers who want repository-wide agentic refactoring inside a purpose-built editor should look at Cursor (from $16/mo billed annually) or Claude Code (from $17/mo billed annually). Teams that need on-premise or offline operation should look at a self-hostable agent, since Copilot runs as a hosted service and authenticates against GitHub. Teams wanting a natively built JetBrains experience rather than a plugin layer should look at JetBrains AI Assistant (from $8.33/mo billed annually). Current state Q3 2026: GitHub expresses AI Credit allowances in dollar values rather than numeric counts. The published model catalog spans Anthropic Claude (Haiku, Sonnet, Opus and Fable), OpenAI GPT-5 series, Google Gemini, xAI Grok, Moonshot Kimi and Microsoft MAI, plus Raptor mini, and it changes often enough that any fixed list dates quickly. Third-party agent delegation to Claude Code and OpenAI Codex is in public preview on Pro+ and Max. The GitHub Copilot Trust Center, scoped to Copilot rather than to GitHub Enterprise Cloud, publishes SOC 2 Type II, ISO 27001 and ISO/IEC 42001:2023, the last certified in March 2026. SOC 1, SOC 3, CSA STAR Level 2, TISAX and PCI DSS are also published there but sit outside this directory's approved certification list. GitHub Copilot holds no FedRAMP authorization of its own. GDPR support is documented on the product page FAQ, which answers yes and grounds it in a Data Protection Agreement between GitHub and the customer. Images sent to Copilot Vision are not used to train or fine-tune generative AI models.