What Sets Them Apart
Jean and Conductor.build both target the same modern problem — running multiple AI coding agents across parallel git worktrees without losing your mind in tab chaos — but they get there from different starting points. Conductor.build is a free macOS-only app with a web-based dashboard angle and a Claude-and-Codex subscription requirement; Jean is an Apache 2.0 native Tauri desktop app with auto-detection for whichever CLIs you already have installed. The split between paid backend dependence and open-source local-first architecture is the most consequential difference, but the day-to-day workflow gap shows up first.
Jean and Conductor.build at a Glance
Jean is built by the coolLabs team (the group behind Coolify), released under Apache 2.0, and runs as a Tauri desktop app on macOS, Windows, and Linux. It detects existing Claude CLI, Codex CLI, Cursor CLI, and OpenCode installations and treats them as first-class backends, so you can mix providers within a single project. The execution model has explicit Plan, Build, and Yolo modes with plan approval, plus Magic Commands for issue investigation, PR drafting, and merge conflict resolution. Headless mode plus token-auth HTTP exposes the same UI over Tailscale for remote access.
Conductor.build is a free desktop app (free in price, not in license) that focuses on monitoring multiple agent sessions through a real-time dashboard. It is currently macOS-only with a Windows waitlist, and it requires you to bring your own Claude or Codex subscription or API key — the app itself does not include the agent. The product positioning emphasizes team-level visibility into agent productivity: who is assigned what, how much each session is consuming, and where outputs need review.
Both apps treat git worktrees as the unit of work and both support multi-agent sessions in parallel. The shared category is real, and the alternatives list on each side correctly cross-references the other. Where the products diverge is in scope: Conductor.build leans toward team-lead oversight of paid agents, while Jean leans toward the individual developer who wants every CLI on their machine in one window.
CLI Backends and Provider Flexibility
Jean's auto-detection covers Claude CLI, Codex CLI, Cursor CLI, and OpenCode without any per-CLI account setup inside Jean itself — if your terminal can run them, Jean shows them in the model picker with the full list of supported variants (Opus 4.5, 4.6, the 1M-context build, Sonnet 4.6, Haiku, plus Codex tiers). This makes it straightforward to keep a Claude-led plan-mode session in one tab and a Codex-led code review in another, or to keep a cheap Haiku Yolo loop running on a sandbox worktree.
Conductor.build is narrower on backends, oriented around Claude and Codex subscriptions or API keys that you bring to the app. The strength of this model is uniform billing visibility — Conductor can show consumption across sessions in a way that bring-your-own-CLI tools cannot — but the trade-off is that you cannot, for example, slot OpenCode into a comparison run alongside Claude without leaving the app.