Can I use Claude models inside GitHub Copilot?

Yes. GitHub offers an Anthropic Claude coding agent, in public preview, that uses the Claude Agent SDK and can be powered by your existing Copilot subscription on any paid plan. You assign it an issue or a prompt, and it makes the changes and opens a pull request for your review, alongside Copilot's own cloud agent.

Is there a free version of Claude Code or GitHub Copilot?

GitHub Copilot Free gives limited access to selected features at no cost, and verified students can use the Copilot Student plan. Claude Code isn't included in Anthropic's Free plan; it comes with paid Claude plans such as Pro, Max, Team and Enterprise, or through an Anthropic Console account billed for API usage.

Does Claude Code work in VS Code and JetBrains IDEs?

Yes. Anthropic provides a VS Code extension, which also installs in Cursor, with inline diffs and plan review, and a plugin for IntelliJ IDEA, PyCharm, WebStorm and other JetBrains IDEs. The JetBrains plugin needs the Claude Code command-line tool installed separately. A desktop app and a browser version are also available.

Is it safe to let an AI coding agent run commands?

It can be, with the right settings. Use permission rules that deny risky commands and secret files, keep approval prompts on for untrusted repositories, run risky work in containers or virtual machines, and require human review before anything merges. Both vendors document controls for this, but configuring them is your team's job.

Which is better for a team that already uses GitHub Enterprise?

GitHub Copilot fits most naturally, because it adds AI to issues, pull requests and code review with centralized policies and audit logs. That doesn't rule out Claude Code: it opens pull requests on GitHub and runs in GitHub Actions, and some teams use both. Trial both on real backlog tasks before committing seats.

Claude Code vs GitHub Copilot: Choosing an AI Coding Assistant

Choose GitHub Copilot if you want AI woven through your editor and GitHub, from inline suggestions to a cloud agent that opens pull requests, with a free tier to start. Choose Claude Code if you want an agent that works across your whole codebase from the terminal, IDE, desktop app or browser. Many teams use both, and GitHub Copilot can even run Claude models.

At a glance (as of October 2026)

Question	GitHub Copilot	Claude Code
Made by	GitHub (Microsoft)	Anthropic
Core idea	Assistant across IDE and GitHub	Agent that edits and runs code
Where it runs	IDEs, GitHub.com, CLI, mobile, desktop app	Terminal, VS Code, JetBrains, desktop, web
Models	A choice from several providers	Claude models
Free option	Copilot Free (limited)	No; paid Claude plan or Console account

How GitHub Copilot works

GitHub's features page splits Copilot into assistive and agentic features.

Assistive features

Inline suggestions as you type in supported IDEs, plus next edit suggestions in VS Code, Xcode and Eclipse that predict where your next change belongs.
Copilot Chat in supported IDEs, on GitHub.com, in GitHub Mobile and in Windows Terminal.
Pull request summaries and commit message help in GitHub Desktop.

Agentic features

Agent mode in the IDE, where Copilot works through a task in your local workspace.
Copilot CLI for working in the terminal, and a GitHub Copilot app for agent-driven development on the desktop.
Copilot cloud agent, which researches a repository, plans and makes changes on a branch, then hands you a pull request to review.
Copilot code review for AI review suggestions on pull requests.
Third-party coding agents (public preview), including Anthropic Claude and OpenAI Codex, which GitHub says are available on all paid Copilot plans and can be powered by your existing Copilot subscription.

You can shape Copilot with custom instructions, prompt files, agent skills, MCP servers and custom agents. Business and Enterprise plans add policy management, audit logs and file exclusions for administrators.

GitHub Copilot plans

GitHub's plans page lists these US prices as of October 2026: Free, Pro at $10 per month, Pro+ at $39, Max at $100, Business at $19 per seat and Enterprise at $39 per seat, plus a free Student plan for verified students. Usage is now measured in GitHub AI Credits, with an allowance included in each paid plan. GitHub states that code completions and next edit suggestions are not billed in AI credits and remain unlimited on paid plans.

How Claude Code works

Anthropic's overview describes Claude Code as an agentic coding tool that reads your codebase, edits files, runs commands and integrates with your development tools. The emphasis is on delegating whole tasks rather than completing lines.

Where it runs

Terminal CLI on macOS, Linux, WSL and Windows.
VS Code (and Cursor) extension, and a JetBrains plugin that uses the CLI.
Desktop app for macOS and Windows, with Linux in beta on Ubuntu and Debian.
Web at claude.ai/code and in the Claude mobile apps, for tasks that run in the cloud.
Slack, where mentioning Claude with a bug report can produce a pull request.

What it can do

Claude Code works directly with git: staging changes, writing commit messages, creating branches and opening pull requests. It can run in CI through GitHub Actions or GitLab CI/CD, connect to outside tools through MCP, and run subagents in parallel. Teams customize it with a CLAUDE.md file of project instructions (it can also read an existing AGENTS.md), skills for repeatable workflows, and hooks that run shell commands before or after its actions, such as formatting after every edit. It is also scriptable: claude -p runs a single prompt, so you can pipe logs or diffs into it.

Claude Code plans

Most Claude Code surfaces require a Claude subscription that includes it or an Anthropic Console account, and the terminal, VS Code and JetBrains versions also support third-party providers. Anthropic's pricing page lists, as of October 2026, Pro at $20 per month billed monthly ($17 per month billed annually), Max from $100 per month, and Team and Enterprise priced per seat. The Free plan does not include Claude Code.

Where each one fits

Both tools now have agents, CLIs, MCP support and project instructions, so the differences are about emphasis and workflow, not checkboxes.

GitHub Copilot is the natural fit when:

Your code lives on GitHub and you want AI inside issues, pull requests and code review, not just the editor.
Developers value fast inline completions while they type, across VS Code, Visual Studio, JetBrains, Xcode or Eclipse.
You want one vendor relationship with centralized seat management and policies for a large organization.
You want a free tier for people to try before you buy seats.

Claude Code is the natural fit when:

Developers prefer to describe a task and review the result, especially multi-file changes, refactors and test writing.
Your team lives in the terminal and wants an agent that scripts into existing pipelines.
You host code outside GitHub (Claude Code supports GitLab CI/CD as well) or want the same agent in the terminal, IDE, desktop and web.
You want deterministic guardrails through hooks and fine-grained permission rules checked into the repository.

It is not strictly either/or. GitHub's Anthropic Claude coding agent runs Claude models inside GitHub on a Copilot subscription. Some teams keep Copilot for completions and review and add Claude Code for larger delegated tasks.

Security and review practices for both

An AI coding agent can run commands with your permissions. Treat it like a new team member with write access: useful, fast and in need of review.

What each tool provides

Claude Code uses permission modes. According to its security documentation, Auto mode is now the starting mode for interactive terminal and VS Code sessions: a separate classifier model reviews actions and blocks those it judges unsafe, while your own ask and deny rules still apply. Manual mode starts read-only and asks before edits and commands. Optional sandboxing isolates Bash commands at the filesystem and network level. Rules live in settings files, and deny rules are evaluated first, then ask, then allow. A project .claude/settings.json might include:

{
  "permissions": {
    "allow": ["Bash(npm run *)"],
    "deny": ["Bash(git push *)", "Read(./.env)", "Read(./secrets/**)"]
  }
}

GitHub Copilot cloud agent has protections GitHub lists in its risks and mitigations page: only users with write access can trigger it; it pushes to a single branch and stays subject to branch protections; it can't approve or merge its own pull requests, and the person who requested the work can't approve it either; Actions workflows don't run until a user with write access approves them, unless the owner allows automatic runs; internet access is restricted; and hidden characters, such as text in HTML comments, are filtered from input. Generated code is checked with CodeQL, the GitHub Advisory Database and secret scanning.

Practices to adopt either way

Protect your main branch with required reviews and status checks. Agents should propose; humans merge.
Never auto-approve everything on untrusted code. Run risky work in a container, dev container or VM; Anthropic recommends VMs especially when an agent interacts with external web services.
Keep secrets out of reach. Deny reads of .env files and key folders, and use short-lived credentials.
Treat issues, web pages and dependencies as untrusted input. They are a route for prompt injection, so be careful what you let an agent read while it can also run commands.
Monitor usage. Claude Code supports OpenTelemetry metrics; Copilot Business and Enterprise provide audit logs and usage data.

How to trial both fairly

Skip public benchmarks and test on your own work, in a short, time-boxed trial such as two weeks.

Pick five real tasks from your backlog: a bug fix, a test-writing task, a small feature, a refactor and a documentation update.
Give each tool the same context. Write project instructions for both, covering build commands, conventions and forbidden areas.
Use the same starting branch and the same reviewer for each attempt.
Record what matters to you: did it produce a mergeable pull request, how many review rounds it needed, whether it followed conventions, whether it touched files it shouldn't, and how much plan usage it consumed.
Ask the developers. Fit with daily habits often decides adoption more than any single result.
Include security in the trial. Have someone confirm your permission settings and branch rules actually block what they should.

How CertTulen Academy can help

If your team chooses Claude Code, or wants to evaluate it properly, Claude Code in Action is a two-day, hands-on course; the course page has the full outline. The security and review practices above apply equally to GitHub Copilot. CertTulen delivers live, instructor-led online classes and can run a private cohort for your team. Contact us for a batch that suits your time zone.