GitHub Copilot

What it does

GitHub Copilot is GitHub’s AI pair programmer and coding agent product. It started as an autocomplete and code suggestion tool, but as of June 2026 Copilot has expanded into a much broader developer platform: inline completions, next edit suggestions, IDE chat, agent mode, cloud agent, code review, CLI, GitHub Mobile, pull request summaries, issue workflows, and MCP connections.

Copilot’s biggest differentiator is that it is embedded directly into the GitHub ecosystem. Repositories, issues, pull requests, Actions, security alerts, branches, commits, and organization policies are native parts of the Copilot workflow. This makes Copilot more than a “coding chatbot”; it is a developer productivity and agentic software workflow layer running on GitHub.

As of June 2026, Copilot supports multiple model providers. OpenAI models, Anthropic Claude models, Google Gemini models, Microsoft models, and Copilot-specific fine-tuned models can be used inside the same product. Model access can vary by plan, mode, organization policy, and region.

Models

GPT-5.5 — Frontier OpenAI model for advanced chat, reasoning, editing, and agent workflows.

GPT-5.3-Codex — OpenAI’s coding-focused model for agent mode, code generation, debugging, and software engineering tasks.

Claude Sonnet 4.6 — Anthropic model for balanced coding, refactoring, explanations, and agentic edits.

Claude Opus 4.8 — High-intelligence Anthropic model for complex coding tasks, deep reasoning, and large refactors.

Gemini 3.5 Flash — Google model for fast coding assistance, chat, and agent workflows.

MAI-Code-1-Flash — Microsoft coding model optimized for fast Copilot coding assistance and agent workflows.

Copilot is not a single-model product. Users can choose between different models in supported plans and modes, or use auto model selection.

Copilot model costs are not exposed to users as a classic input/output token API table. As of June 2026, features such as chat, agent mode, code review, cloud agent, CLI, and Copilot Apps consume GitHub AI Credits. 1 AI Credit = $0.01 USD. Code completions and next edit suggestions remain included in subscriptions and do not consume credits.

Pricing

• Free ($0/mo) — limited chat/agent, 2,000 completions per month, IDE, CLI, GitHub Mobile access
• Pro ($10/mo) — unlimited completions, $10 base + $5 flex = $15 monthly AI Credits, IDE chat, agent mode, code review, CLI
• Pro+ ($39/mo) — premium model access, $39 base + $31 flex = $70 monthly AI Credits, more agent/chat
• Max ($100/mo) — highest individual tier, $100 base + $100 flex = $200 monthly AI Credits, priority access
• Business ($19/seat/mo) — $19 pooled AI Credits per seat, org management, policy controls
• Enterprise ($39/seat/mo) — $39 pooled AI Credits per seat, Business + enterprise controls, knowledge bases

As of June 2026, Copilot has moved from the premium request model to GitHub AI Credits. Pro/Pro+/Max plans have a “base + flex” allotment structure: base credits drain first, flex credits are used for overflow. Annual plans are being retired; existing annual subscribers continue with PRU (premium request unit) until expiration.

Important note: New sign-ups for Pro/Pro+/Max plans have been paused since April 20, 2026. Existing users can still upgrade.

Capabilities

• Inline code completion
• Next edit suggestions
• Copilot Chat inside IDEs
• Chat on GitHub.com and GitHub Mobile
• Windows Terminal and Copilot CLI
• Agent mode for local multi-step coding tasks
• Copilot cloud agent for branch-based autonomous work on GitHub
• Pull request summaries and code review
• Integration with issues, PRs, security alerts, and Actions workflows
• MCP connections to external tools and resources
• Multi-model selection (OpenAI, Anthropic, Google, Microsoft)
• Organization policy, content exclusion, and admin controls

Strengths

• Native access to GitHub repository, issue, PR, branch, commit, and Actions context is extremely strong
• Works across a wide range of developer environments: VS Code, Visual Studio, JetBrains, Xcode, Neovim, Eclipse, Zed, CLI, GitHub, and mobile
• Not limited to one model; supports OpenAI, Anthropic, Google, Microsoft, and Copilot-tuned models
• Very strong for everyday developer productivity: autocomplete, explanations, refactors, tests, edits, and PR help
• Copilot cloud agent can turn simple issues and repository tasks into branch/pull request workflows
• Business and Enterprise plans provide mature policy, content exclusion, pooled usage, and admin controls
• Reduces context switching for teams already working on GitHub

Weaknesses

• The June 2026 AI Credits transition makes cost modeling more complex than the older simple subscription model
• Code completions remain free; everything else (chat, agent, review, cloud agent) consumes credits
• Model access can vary by plan, mode, region, and organization policy
• Code produced by cloud agent and agent mode must still be carefully reviewed by humans
• New sign-ups for Pro/Pro+/Max plans have been paused since April 20, 2026
• Free plan is limited for heavy professional use
• Behavior can differ across surfaces: IDE chat, agent mode, cloud agent, CLI, and GitHub.com do not provide identical experiences
• Its advantage is weaker for teams that do not use GitHub as their main development platform

Ecosystem

The GitHub Copilot ecosystem is not a classic chatbot ecosystem; it is a GitHub-native developer workflow ecosystem.

IDE Integrations bring Copilot into VS Code, Visual Studio, JetBrains IDEs, Xcode, Neovim, Eclipse, Zed, SQL Server Management Studio, and other supported environments. This is the main layer for autocomplete, chat, edits, and agent mode.

Copilot Cloud Agent is the autonomous coding agent layer running on GitHub. It can research repositories, create implementation plans, make code changes on branches, run tests/lints, produce commits, and create pull requests when needed. Unlike IDE agent mode, it works in a GitHub Actions-powered ephemeral cloud development environment.

Agent Mode is the more local agentic mode inside IDEs. It can perform multi-step coding tasks in the user’s development environment. When combined with MCP, it can connect to external resources and tools to complete stronger agentic loops.

MCP Support allows Copilot agent mode to connect to external tools and resources. Internal docs, APIs, databases, issue trackers, or custom services can be brought into Copilot workflows.

Copilot CLI enables natural-language coding, debugging, and command-line help in the terminal. It is positioned as a command-line coding assistant integrated with GitHub workflows.

Copilot Code Review provides AI-assisted review for code selections and pull requests. As of June 2026 it moved to an agentic architecture running on GitHub Actions, meaning PR reviews also consume Actions minutes.

GitHub Platform is Copilot’s biggest ecosystem advantage. Repository, issue, pull request, Actions, security alert, organization policy, audit, and admin controls turn Copilot from a standalone AI coding tool into a developer automation layer embedded inside GitHub.