Claude Code vs GitHub Copilot: agent mode, cloud agent, CLI and pricing compared

What is the difference between Claude Code and GitHub Copilot?

Claude Code is Anthropic's coding agent; GitHub Copilot is GitHub's AI assistant, which now includes several agents. Claude Code started in the terminal and you mostly hand it tasks. Copilot started as autocomplete in the editor and now spans inline suggestions, agent mode in the IDE, a terminal CLI and a cloud agent that works from GitHub issues.

So the choice is partly about models and partly about where you want the work to happen. Copilot lives inside GitHub and your IDE, and gives you models from several providers. Claude Code gives you Anthropic's agent and Claude models in every surface it runs on, from your terminal to the browser.

How does Claude Code work?

Claude Code is "an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools." It runs in the terminal, as a VS Code extension (which also installs in Cursor), as a JetBrains plugin, in a desktop app and on the web. Each surface uses the same engine, so your CLAUDE.md, settings and MCP servers carry across.

A typical session: run claude in your project, describe a task, review what it proposes, and commit. For scripts and CI, claude -p "query" runs one task and exits. For GitHub, Claude Code GitHub Actions responds to @claude mentions in issues and pull requests, and /install-github-app sets it up from a session.

Claude Code uses Anthropic's Claude models (Opus, Sonnet and Haiku, with Fable available for the hardest tasks). See the setup guide to install it.

How does GitHub Copilot work?

GitHub Copilot is "an AI assistant that helps you write, understand, and ship software." The agentic parts come in four places:

• Agent mode in your IDE. In VS Code, Visual Studio, JetBrains IDEs and other supported editors, the agent finds relevant code, edits files, runs commands and checks its work while you review and redirect it.
• Copilot cloud agent. On GitHub, you delegate a task from the Agents tab, an issue, a pull request or Copilot Chat. The cloud agent works in "an ephemeral cloud development environment," makes changes on a branch and can open a pull request.
• Copilot CLI. The copilot command gives you an agent in the terminal, with a plan mode, a -p flag for one-shot prompts, and /model to switch models.
• The GitHub Copilot app, a desktop app for running several agent sessions in parallel, locally or in the cloud.

Copilot also has inline code suggestions, code review on pull requests, automations that run the cloud agent on a schedule or on repository events, and support for third-party coding agents, including an Anthropic Claude agent and OpenAI Codex, both in public preview.

Claude Code vs Copilot side by side

How do the CLIs compare?

The two terminal tools look similar on the surface. Both start an interactive session, both have a plan mode, and both take -p for a single prompt that runs and exits.

Installing Copilot CLI, per GitHub's quickstart, is one of these:

npm install -g @github/copilot      # needs Node.js 22 or later
brew install --cask copilot-cli     # macOS and Linux
winget install GitHub.Copilot       # Windows

Then run copilot and sign in with /login. Claude Code installs with a native script, Homebrew (brew install --cask claude-code) or WinGet (winget install Anthropic.ClaudeCode), and you sign in with a Claude account.

Two differences stand out. Copilot CLI can switch between GitHub-hosted models from several providers, and it can point at your own provider (an OpenAI-compatible endpoint, Azure OpenAI or Anthropic) through environment variables. Claude Code stays on Claude models but can reach them through Amazon Bedrock, Google Cloud and Microsoft Foundry.

How do they handle safety and approvals?

Claude Code has six permission modes: default, acceptEdits, plan, auto, dontAsk and bypassPermissions, plus allow and deny rules. Its OS-level sandbox for shell commands is off by default; /sandbox turns it on, on macOS, Linux and WSL2.

Copilot CLI asks before it changes files, and you loosen that with flags such as --allow-tool='shell(git)' or --deny-tool. GitHub warns that --allow-all-tools gives Copilot the same access to your files and shell as you have. /sandbox enable turns on local sandboxing, and copilot --cloud starts a session in a cloud sandbox, which is in public preview.

The cloud agents are isolated by design: Copilot cloud agent works in GitHub's ephemeral environment and returns a pull request, and Claude Code cloud sessions run on infrastructure Anthropic manages, started from the web, the desktop app or claude --cloud.

How do instructions and skills carry over?

This is where the two overlap more than you might expect. GitHub's custom instructions support table lists AGENTS.md, CLAUDE.md and GEMINI.md as agent instructions for Copilot cloud agent and Copilot CLI. For Copilot Chat in VS Code, the table lists AGENTS.md. Copilot's own files are .github/copilot-instructions.md and path-specific .github/instructions/*.instructions.md.

Claude Code reads CLAUDE.md, and reads AGENTS.md when there is no CLAUDE.md. The simplest shared setup is an AGENTS.md file, with @AGENTS.md at the top of CLAUDE.md if you keep one.

Skills line up too. GitHub's agent skills page says Copilot supports project skills in .github/skills, .claude/skills or .agents/skills. So skills you keep in .claude/skills/ for Claude Code are also picked up by Copilot.

How much do Claude Code and Copilot cost?

Prices below are as of October 2026.

• GitHub Copilot (plans): Free, with limited agent use and 2,000 completions a month; Pro at $10 a month with 1,500 AI credits; Pro+ at $39 with 7,000; Max at $100 with 20,000; Business at $19 per seat and Enterprise at $39 per seat. All plans include Copilot CLI, cloud agent is not on Free, and code completions are unlimited on paid plans. AI credits are spent by tokens, so pricier models use more.
• Claude Code (pricing): included in Pro at $20 a month ($17 a month billed annually), Max from $100 a month, and Team seats from $20 per seat per month billed annually. The free Claude plan does not include it. A Claude Console account bills per token instead. The Claude Code pricing guide covers limits.

Which should you pick?

GitHub Copilot tends to fit when:

• You want inline suggestions as you type, alongside an agent.
• Your work starts in GitHub issues and you want an agent to turn them into pull requests.
• You want to switch between models from several providers, or bring your own.
• Your company already buys Copilot Business or Enterprise.

Claude Code tends to fit when:

• You prefer Claude models and want Anthropic's own agent, not a wrapper.
• You want a choice of permission modes, from asking about every action to auto mode.
• You work in the terminal and want the same agent in your IDE, desktop and browser.
• You already pay for a Claude plan.

You can also use both, for example Copilot for suggestions and code review, and Claude Code for larger tasks. Since Copilot reads .claude/skills/ and AGENTS.md, one repo setup can serve both.

Keep your prompts and skills in one place

Whichever agent you use, the skills and prompts you refine are worth keeping somewhere you own. Kipt stores them as Markdown in a private GitHub repo and installs every skill into Claude Code or Codex with one npx skills add command. To turn a prompt you keep pasting into a skill folder, try the free SKILL.md generator.