Run coding agents in local Mac workspaces or shared cloud sandboxes with Conductor.
Conductor provides a workspace for running several coding agents in parallel. Developers can work locally on a Mac or use cloud sandboxes on a paid plan. The product runs first-party coding agents under the hood, including Claude Code, Codex, Cursor and OpenCode, and lets users bring their own subscriptions or API keys for model access.
The free plan supports parallel agents in local workspaces. Cloud workspaces keep repositories and dependencies in isolated environments so an agent can continue working when the desktop application is closed. That distinction is important for longer tasks: a local session depends on the user’s machine remaining available, while a cloud session runs on managed infrastructure.
Conductor’s multiplayer feature allows teammates to share a workspace link, see who is active and prompt agents together. The interface also supports reviewing changes and editing files in diffs. Parallel execution can help separate independent tasks, but each task still needs a clear scope so that simultaneous work does not create conflicting changes or accidental permission expansion.
Pro is listed at $50 per month and adds cloud workspaces, multiplayer and the Conductor API. Teams is listed at $60 per user per month and adds organisation management and centralised billing. Enterprise is quoted separately and includes additional identity, support and contractual controls. The reviewed pricing page describes team plans as invitation-based.
The mobile application is advertised as coming soon, not as an available native app. Cloud compute terms may also evolve: the company says usage-based cloud charges are planned. Model subscriptions and provider charges should be considered separately from the Conductor plan rather than assumed to be included.
Local and cloud sessions have different data handling. Conductor explains that local session content goes to the connected model provider and is stored on the device, while cloud inputs and outputs must also be stored on its servers. Begin with a bounded repository task, inspect the diff and test results, and use the team’s normal review process before merging changes.
