Can I run Claude Code on Windows without WSL?

Yes. Claude Code runs natively on Windows since early 2026. You need Node.js 18+ and git installed. WSL is optional but useful for projects that depend on Linux toolchains.

Should I use WSL or native Windows for AI coding agents?

Use native Windows if your project builds and runs natively on Windows. Use WSL if your project requires Linux tools (Docker, bash scripts, Linux-specific dependencies). Pane works with both and can reach repositories inside WSL from the Windows desktop.

Does Pane support Windows ARM64?

Yes. Pane ships separate builds for Windows x64 and Windows ARM64 (Surface Pro X, Snapdragon laptops). Both are fully supported.

How do I install Codex on Windows?

Run 'npm install -g @openai/codex' in your terminal. Codex requires Node.js 22+ and an OpenAI API key set as OPENAI_API_KEY in your environment.

How do I install Aider on Windows?

Run 'pipx install aider-chat' or 'pip install aider-chat'. Aider requires Python 3.9+ and an API key for your preferred model provider.

Can I run multiple agents on Windows at the same time?

Yes. Each agent runs in its own terminal. The challenge is managing terminals, branches, and worktrees across multiple sessions. Pane automates this by giving each agent its own worktree and terminal in a single desktop app.

What if I get 'claude is not recognized' after installing?

This usually means the npm global bin directory is not in your PATH. Run 'npm config get prefix' to find it, then add the bin subdirectory to your system PATH. Restart your terminal after updating PATH.

Do I need tmux on Windows?

No. tmux does not run natively on Windows. Tools that require tmux (like Claude Squad) need WSL or a Linux environment. Pane does not require tmux and includes its own terminal emulator.

How do I use Pane with a project inside WSL?

Launch Pane from Windows, then open your WSL repository path. Pane detects the WSL filesystem and handles path translation automatically when creating worktrees.

← back to runpane.com

AI coding agents on Windows: the complete setup guide

This guide takes you from a fresh Windows machine to running multiple AI coding agents in parallel. It covers installing the prerequisites, choosing between native Windows and WSL, setting up agents, and managing them with Pane. If you already have git and Node.js installed, skip to installing agents.

prerequisites

Every AI coding agent needs git, and most need Node.js or Python. If you already have these, you are set. If not, here is the fastest path:

git

Download from git-scm.com or install with winget:

winget install --id Git.Git -e --source winget

During installation, accept the default settings. Make sure "Git from the command line and also from 3rd-party software" is selected so git is available in PowerShell.

node.js

Claude Code and Codex both need Node.js. Download the LTS version from nodejs.org or install with winget:

winget install --id OpenJS.NodeJS.LTS -e

After installation, open a new terminal and verify with node --version. You need v18 or higher for Claude Code, v22 or higher for Codex.

python (optional)

Aider and some other agents run on Python. Install from the Python website or with winget:

winget install --id Python.Python.3.12 -e

Check "Add Python to PATH" during installation. You need Python 3.9 or higher for Aider.

native windows vs WSL

Windows Subsystem for Linux (WSL) gives you a full Linux environment inside Windows. Some developers put all their projects in WSL. Others stay native. Both work with AI coding agents. Here is when to use each:

	native Windows	WSL
best for	.NET, Unity, PowerShell, projects that build natively on Windows	Node.js, Python, Docker, bash-heavy projects, Linux toolchains
file system speed	fast on NTFS	fast inside WSL, slow if accessing Windows files from WSL
git worktrees	work, but watch for path length limits	work, keep worktrees inside WSL filesystem
Pane	native installer, runs directly	runs on Windows, reaches into WSL repos

If you are unsure, start with native Windows. You can always set up WSL later. To install WSL:

wsl --install

installing AI coding agents

You only need to install the agents you plan to use. Here are the most common ones and their Windows install commands:

claude code

Anthropic's CLI agent. Best at large-scope refactors and understanding complex codebases. Requires Node.js 18+ and an Anthropic API key or Claude Pro/Max subscription.

npm install -g @anthropic-ai/claude-code
claude

codex

OpenAI's CLI agent. Fast at targeted edits and test generation. Requires Node.js 22+ and an OpenAI API key.

npm install -g @openai/codex
codex

aider

Open-source pair programming agent. Supports multiple model providers. Requires Python 3.9+ and an API key.

pipx install aider-chat
aider

opencode

Terminal-native coding agent. Supports Anthropic and OpenAI models.

npm install -g opencode-ai
opencode

After installing, verify each agent works by running its command in a new terminal window. If you get "not recognized as a command," close and reopen your terminal so PATH changes take effect.

installing pane

Pane is the workspace where your agents live. It handles worktrees, terminals, diffs, and git workflow so you do not have to manage those manually. Install with the guided CLI:

npx --yes runpane@latest

Or use the PowerShell installer directly:

irm https://runpane.com/install.ps1 | iex

Both paths install the same Pane desktop app. The npm path gives you a guided setup flow. The PowerShell path is a direct download-and-install. After installation, launch Pane from the Start menu or type pane in any terminal.

your first parallel session

Once Pane is running, open a repository. Then create two panes, each for a different task:

→ Pane 1: Claude Code refactoring the authentication module.
→ Pane 2: Codex writing tests for the payment API.

Each pane gets its own git worktree and terminal. You can switch between them with keyboard shortcuts, see the diff for whichever pane is focused, and review/commit/push from the same interface. When you are done with a task, close the pane and the worktree is cleaned up.

For more on the parallel workflow, read How to Run Multiple AI Agents in Parallel.

common windows issues

"claude is not recognized as a cmdlet"

The npm global bin directory is not in your PATH. Run npm config get prefix to find the install location. Add its bin subdirectory to your system PATH. Then open a new terminal window.

NTFS long path errors

Windows has a 260-character path limit by default. Deep node_modules directories can hit this. Fix it with:

git config --system core.longpaths true

Also enable long paths in Windows: Settings → System → For developers → Enable long paths. Or via registry:

reg add "HKLM\SYSTEM\CurrentControlSet\Control\FileSystem" /v LongPathsEnabled /t REG_DWORD /d 1 /f

port conflicts between worktrees

If you run dev servers in multiple worktrees, each one tries to bind to the same port (usually 3000). Pane handles port isolation automatically. If running manually, set the port via environment variables (e.g. PORT=3001 npm run dev).

slow git operations across WSL boundary

If your repo is inside WSL, run git commands from inside WSL too. Accessing WSL files from Windows (or vice versa) is slow because it crosses a filesystem translation layer. Pane handles this correctly, but manual terminal workflows should stay on one side of the boundary.

frequently asked questions

Download*

Windows SmartScreen warningDirect downloads can show a SmartScreen warning while Pane is unsigned. Pane is fully open source, so you can audit the code and build from source yourself.1. Click More info2. Click Run anyway3. Continue the installerThe PowerShell install downloads the official release directly and avoids most browser download friction.npm global install

Paste that in PowerShell.MacLinux