Alternatives · parallel Claude Code workspace for macOS

Open source alternative to Conductor

Conductor runs several Claude Code sessions side by side, each in its own Git worktree, so two agents can work on the same repository without stepping on each other. It is a well-made app and the worktree model is the right one. What it does not do is tell you whether any of that work was correct, and there is no source to read or self-host.

Apache 2.0Runs on your machineAny agent
Conductor licence

Closed source. No public repository.

Platforms

macOS only.

Pricing

Free to use. You pay Anthropic for the agent.

The honest part

Why people replace Conductor

You are not on a Mac

Conductor ships for macOS. If half your team is on Linux, or your work happens on a remote box, the app is not an option for them.

One agent vendor

It is built around Claude Code. That is a reasonable bet, but it means the workspace and the model are bought together, and switching one means leaving the other.

No source to read or run

There is no repository. You cannot audit what it does with your code, patch a bug yourself, or run it inside a network that does not reach the public internet.

Parallel work still lands unreviewed

Running four agents at once produces four diffs. The reading is still yours, and four agents produce diffs faster than one person can read them.

What Aura does instead

The layer above your agents

The same worktree model, on Linux too

Each agent gets an isolated worktree and its own branch. The app is a Tauri build for macOS and Linux, and the CLI runs anywhere Rust runs, including over SSH on a box with no display.

Any agent, one history

Claude Code, Codex, Gemini CLI, opencode and the rest run in the same window against the same repository. The record of what happened does not change when you change model.

Apache 2.0, the whole thing

App, CLI, engine and editor extension are in one public repository under Apache 2.0. Read it, fork it, run it on a machine with no route to the internet.

A verdict on the diff, not just the diff

Aura parses each change into an AST graph, compares it against the goal the session was given, and reports which parts are wired, partly wired, or missing. That is the part parallel agents make urgent.

What you give up

Aura is not better at everything

  • Conductor is more finished as a Claude Code workspace. If Claude Code is the only agent you will ever run, its setup is shorter.
  • Its Linear and GitHub issue pickup is more mature than ours.
  • Aura asks you to care about goals and intent. That is a habit, and habits cost something on day one.
Switching

How the move actually goes

  1. 01Install Aura and open the same repository. Nothing about the repo changes; it stays an ordinary Git repository the whole way through.
  2. 02Point it at the agent you already pay for. Claude Code keeps its own credentials and settings.
  3. 03Start a session with a goal written in a sentence. That sentence is what the proof is checked against later.
  4. 04Run the agents you were running. The worktrees, branches and merges work the way they did.
  5. 05Read the semantic diff instead of the text diff, and check the verdict before you merge.
Questions

Aura and Conductor

Is Aura a drop-in replacement for Conductor?

For the parallel-worktree part, close to it: same model of one agent per worktree, same merge path. Aura adds the review layer and drops the macOS-only constraint. Conductor has the smoother first ten minutes if you only run Claude Code.

Does Aura only work with Claude?

No. Claude Code, Codex CLI, Gemini CLI, opencode, Amp and others run in the same window. The semantic record is the same whichever one produced the change.

Can I self-host it?

Yes. The engine and CLI are local by default, and the collaboration server can run on your own infrastructure. Nothing has to leave your network.

What licence is Aura under?

Apache 2.0, in a single public monorepo covering the desktop app, the CLI, the engine and the editor extension.