Conductor runs Claude Code, Codex and other agents in parallel, but its local workspaces are for your Mac (checked on October 28, 2026). If you work on Windows or Linux, or on more than one of them, Alchemist Coder gives you parallel agents in git worktrees on all three.
Parallel agents, each in its own worktree
- The Arena: a planner drafts a plan in read-only mode, you approve it, and one to four agents (mix Claude Code and Codex, and models) solve the task in parallel, each in its own git worktree under
.alchemist/worktrees. - Compare, then merge: their changes, diffs, tests, time and cost side by side. Merge the winner with a squash or a merge commit; your checkout is untouched until then.
- An organization: a coordinator hands assignments to agents that can each work in their own copy, and you apply their work after a review.
- Every agent in view: a live tree of agents and subagents, and a status bar that says who needs you.
Installers for each system
A DMG for Apple Silicon or Intel Macs, a setup.exe for Windows, and an AppImage or .deb for Linux. Every installer is installed and opened on a real machine of its platform before a release.
Where Conductor is ahead
- It runs more agents (Cursor and OpenCode as well), and its free plan uses your own subscriptions and keys.
- Its paid plans add cloud workspaces, multiplayer and a mobile app. Alchemist Coder has no cloud.
What the free edition does not do
The free Community edition runs Claude Code and Codex on local models only, through Ollama or LM Studio. Running agents on your Claude or ChatGPT subscription, or on API keys, belongs to Pro, which is not on sale yet. Searching and reading your history works whichever plan you use.
The builds are not signed by Apple or Microsoft yet, so macOS and Windows warn you the first time. The download section explains the two clicks it takes.
Questions
Does Conductor work on Windows or Linux?
Its pricing page describes local workspaces “on your Mac”, and its cloud workspaces run on Amazon Linux machines (checked on October 28, 2026).
How many agents can run at once in Alchemist Coder?
The Arena runs up to four agents on the same task, each in its own git worktree. Your organization can have more agents working on different assignments.
More guides
Facts about other tools were checked on October 28, 2026, at the links in the text. Alchemist Coder’s source is on GitHub.