claude code on windows + wsl: manage terminal agents
TL;DR
Claude Code is Anthropic's CLI coding agent, and running several at once means juggling terminal tabs, branches and reviews. Pane gives each Claude Code session its own pane with a worktree, a real terminal, a diff view and git buttons, natively on Windows, inside WSL, on macOS and on Linux. It runs the real claude CLI, so your subscription, slash commands and settings work as usual. If you only ever run one session, Windows Terminal or tmux is enough.
Pane is a desktop agent manager for developers who like terminal agents but hate managing ten terminal tabs. Run Claude Code, Codex, Cursor, Aider, Goose, or any CLI agent in parallel, each with its own workspace and git review loop.
what is claude code?
Claude Code is Anthropic's official CLI agent. It takes a task in natural language, reads your codebase, and executes multi-file edits autonomously - refactoring, debugging, or implementing features end-to-end. It runs best with an Anthropic subscription (Claude Max or API access).
Claude Code is particularly strong at large-scope refactors and agentic loops that span many files. Its source is published in the anthropics/claude-code repository under Anthropic's terms of use. Engineers who need long-running autonomous workflows - not just autocomplete - use it as their primary agent.
I run Claude Code as my main coding agent and use Pane to keep separate Claude Code sessions isolated by branch - one per active feature or refactor - so diffs don't bleed across tasks.
the problem is not vscode
VS Code terminals are good terminals. The problem starts when Claude Code is fixing one feature, Codex is debugging another, a test watcher is running, and every pane needs a branch, logs, and a diff. At that point you are not managing terminals. You are managing agent work.
why windows and wsl change the answer
Most agent-manager advice assumes macOS or tmux. If you are on Windows or WSL, that advice runs out the moment you try to install it. Pane is a native desktop app on Windows, Mac, and Linux, so the workflow is the same wherever you happen to code.
what pane does
- starts each agent in a separate session
- uses git worktrees so agents do not collide
- keeps terminal output, diff review, editor, logs, and git actions together
- lets Claude Code, Codex and Cursor run side by side, and gives Claude Code Pane's MCP tools so it can start and check on other agents
when windows terminal or tmux is enough
If all you need is two shells, use Windows Terminal panes, tmux, or Zellij. They are good tools. Pane is for the next layer: named agent sessions, worktree isolation, diff review, and the decision of what actually ships.
Bring your chats to Pane
Pane reads your own Claude Code folder, ~/.claude or $CLAUDE_CONFIG_DIR, and lists the chats you used in the last 30 days. Each opens in a pane on the repo it ran in, and Claude Code resumes it by its own session id, so the chat picks up where it left off. Several chats in one repo become tabs in one pane.
It skips subagent threads and claude -p or Agent SDK runs. A chat you started in a subfolder of a repo isn't placed by Find sessions: say “migrate to Pane” for those. That also works for a chat that's still running, which it forks onto a new branch while the original keeps going.
To start, open Settings → AI & Agents → Bring sessions into Pane and click Find sessions, or say “migrate to Pane” in any Session, the chat in Pane that runs your agents.
Compare: Pane vs Claude Squad, Pane vs Conductor, Pane vs Crystal.
Background: git worktrees for AI agents, desktop agent managers, tmux agent managers, Pane vs terminal multiplexers.
Last verified against Claude Code's current release. Written by Parsa Khazaeepoul, co-founder of Pane.