GitHub Copilot

GitHub Copilot is the market's oldest commercial AI coding assistant - technical preview launched June 29, 2021, GA on June 21, 2022. Developed by GitHub Inc. (Microsoft subsidiary since 2018 - $7.5B acquisition) in partnership with OpenAI, which delivered the first-generation model (Codex based on GPT-3). Today GitHub Copilot serves over 20 million active users (GitHub Q4 2025 data) and more than 90,000 organisations, including 90% of the Fortune 100. The product is platform-agnostic (the user chooses from a palette of models: Claude, GPT-5.5, Gemini, Grok, GitHub's own model) and positions itself as a neutral developer tool with deliberate selection of the best model for the task.

Base mode: code auto-completion (tab completions, ghost text) in the editor. The model predicts the next lines of code based on context (current file, neighbouring files in the project, entire repo code for Copilot Business and up). Extended with Next Edit Suggestions - the model predicts not only line completion but a sequence of consistent edits across multiple files (e.g. after renaming a function, it points to every place in the project where the import + calls need to be updated).

Chat and agent modes: Copilot Chat - dialogue with the model in an IDE panel (VS Code, JetBrains, Visual Studio, Xcode, Neovim). Model choice per query: Claude Sonnet 4.5/Opus, GPT-5.5, Gemini 2.5, Grok 4.5, GitHub's own model. Agent Mode (since October 2024, GA February 2025) - an interactive agent executing tools (terminal, file edits, tests) in a plan-execute-verify loop under user supervision. Copilot Coding Agent (since May 2025) - an autonomous cloud agent: receives a GitHub Issue, works independently in an isolated environment (GitHub Actions runner), commits changes, and opens a Pull Request. Does not require an open IDE.

Related features on the GitHub platform: Copilot Autofix (since August 2024) - automatic fix suggestions for GitHub Code Scanning alerts (CodeQL, Dependabot, secret scanning). Reduces mean remediation time from days to minutes. Copilot Code Review (GA March 2025) - an AI reviewer leaves inline Pull Request comments like a regular reviewer, flagging bugs, security holes, and code smells. Copilot Extensions (third-party marketplace, GA April 2025) - vendors like Datadog, Sentry, Jira, Docker, MongoDB, Stack Overflow, Azure DevOps add their own tools available from the Copilot chat without leaving the editor.

Positioning and competition: Cursor (Anysphere) - an IDE-first alternative, historically oriented toward Anthropic/Claude, with a well-solved Composer/Agent UX; Claude Code (Anthropic) - terminal-first, quality leader on Sonnet 4.5/Opus; Codex CLI (OpenAI) - OpenAI's terminal-first agent with GPT-5.5; Grok Build (xAI) - terminal-first agent with Grok 4.5, the cheapest option; Windsurf (acquired by Cognition in 2025). GitHub Copilot stands out with: (a) availability inside every IDE instead of a dedicated editor, (b) the deepest integration with the GitHub platform (Issues, Pull Requests, Actions, Code Scanning), (c) enterprise-grade offering (SSO/SAML, SCIM, audit logs, IP indemnity, DPA), (d) the largest selection of selectable models in a single system. Weaknesses: onboarding barriers (from April 2026 new Pro/Pro+/Student sign-ups are temporarily paused), historical concerns about anti-competitive dependence on OpenAI models (now eased by the multi-model palette).