Macaron vs Conductor

At a glance

How each works day to day

Conductor

You add a repository, and each new task gets its own workspace: a git worktree on a fresh branch, with its own terminal and diff. You run several agents at once, review each diff, open a pull request, merge, and archive the workspace. Plan mode asks the agent for a plan you approve before it edits.

Macaron

You open a project in a folder on your Mac or on a Computer, start Claude Code, Codex or Macaron's own agent, and chat. Each turn ends in a card with the files changed and an undo button. Your checks run before and after, Review shows the diff, and the board view lays out every thread. You can open a pull request from the app through the GitHub CLI.

Where code runs and what it can touch

Conductor's docs describe local workspaces as git worktrees. A worktree keeps each task's branch and files apart from the others, but the agent still runs on your Mac as you, with whatever access the agent itself allows. On Pro and above, Conductor Cloud runs sessions in managed cloud sandboxes, and Conductor stores that data on its servers.

Macaron's Computers are Linux machines with no access to your Mac's files, keys or Docker socket, in Docker on your Mac or on a cloud server you connect over SSH. A sandbox copy of a Mac folder works there, and Apply to Mac writes the result back only if it applies cleanly. Deploy and database credentials reach a run only through connectors you switch on.

Review, checks and undo

Conductor's checks are signals about whether a branch is ready to merge: git status, pull request details, CI results, deployments, GitHub comments and todos. Macaron's project checks are your own scripts, run by the system before and after the agent's change, so you can see whether it broke something that was passing. Macaron also makes every turn a git checkpoint you can undo, and can ask a second model to review the diff.

Pricing

Conductor's plans as published on conductor.build/pricing on October 1, 2026.

One Macaron plan; models on your own keys at your provider's price; cloud machines billed by the hour.

Who should pick which

• You want five agents on five branches at once, on your Mac, for free: Conductor.
• You use Cursor's or OpenCode's agent as well as Claude Code and Codex: Conductor.
• Teammates should prompt in the same workspace, or you want a mobile app: Conductor Pro.
• You don't want the agent running on your Mac at all, but still want the code there at the end: Macaron, with a sandbox copy and Apply to Mac.
• You want your tests run before and after every change, not only CI on the pull request: Macaron.

Using both, or switching

Both drive the same Claude Code and Codex CLIs with the same accounts, and both end in ordinary branches. Moving between them means opening the same repository in the other app; there's no project format to convert.

What Macaron doesn't do

No Slack integration, no mobile app, no hosted SLA, no multiplayer, and no visual browser testing of its own beyond the AI end-to-end tests. It also doesn't run Cursor's or OpenCode's agents.