Tom Osman
←[ RETURN_TO_INVENTORY ]

Conductor: Run Coding Agents in Parallel

// August 20, 2026

Conductor homepage
Run a team of coding agents on your Mac

Conductor is a multi-agent coding workspace for running first-party Claude Code, OpenAI Codex, Cursor, and OpenCode agents in parallel. It now supports both its Mac app and cloud workspaces, so agents can work for hours inside isolated environments while you review and steer the results. (Conductor)

What changed

The original Conductor workflow was local and worktree-driven. The current product adds cloud sandboxes, multiplayer workspaces, mobile access, and an API while retaining the desktop workflow. Conductor Cloud provisions an isolated microVM for each agent with repository dependencies ready to go.

Why it's useful

  • Parallel agent execution. Give separate agents focused tasks and let them work at the same time instead of serially sharing one checkout.
  • Isolated environments. Local worktrees and cloud microVMs keep concurrent changes separated and easier to review.
  • Bring your own agents. Use the first-party Claude Code, Codex, Cursor, or OpenCode experience rather than a generic imitation.
  • Multiplayer steering. Share a workspace, see who is active, follow important work, and prompt agents with teammates in real time.
  • Work from more places. Start from the Mac app, use cloud workspaces, or integrate through the Conductor API.

Common plays

  1. Split a feature by boundary. Put UI, API, tests, and documentation in separate workspaces with explicit ownership.
  2. Compare agents. Give Claude Code and OpenAI Codex the same bounded task, then compare the diffs and verification evidence.
  3. Keep a long task off your laptop. Use a cloud workspace when the job needs to continue for hours without tying up your local machine.
  4. Review before integration. Treat each agent's output like a teammate's branch: inspect the diff, run the checks, and merge selectively.

Getting started

  • Get Conductor from conductor.build.
  • Connect a repository and choose a local or cloud workspace.
  • Authenticate the first-party coding agents you already use.
  • Give each agent a concrete task with a clear definition of done.
  • Review the resulting changes and verification output before merging.
Open Conductor