Sculptor: the missing UI for Claude Code

Run parallel Claudes safely in containers. Jump between their environments to instantly test changes. Get suggestions that catch critical issues as you go.

<div align="center"> <table> <tbody> <td align="center"> <a href="https://tryimbue.link/g88kYqM" target="_blank"><strong>Download for Mac (Apple Silicon)</strong></a> </td> <td align="center"> <a href="https://tryimbue.link/XNL0Ken" target="_blank"><strong>Download for Linux</strong></a> </td> <td align="center"> <a href="https://tryimbue.link/IIZhHAd" target="_blank"><strong>Download for Intel Mac [experimental]</strong></a> </td> <td align="center"> <a href="https://discord.gg/sBAVvHPUTE" target="_blank"><strong>Join us on Discord</strong></a> </td> </tbody> </table> </div> <br> <p align="center"> <img width="1312" alt="Sculptor Demo" src="assets/demo.gif" /> </p>

Spin up parallel agents

Each Claude works in its own container. You get safe execution and parallel agents without the hassle of git worktrees. You can even tell them, “Don't come back until my tests pass.”

Instantly pair with any agent

Toggle Pairing Mode to move between any agents' containers. Run, test, and edit each agent's changes from your normal dev environment — no need to switch workspaces.

Merge without conflicts

Merge the changes you like and throw out the ones you don't. Sculptor helps you resolve merge conflicts.

See the power of Sculptor's Pairing Mode in this behind-the-scenes demo 🤓

Here’s our teammate Guinness walking through an earlier build of Sculptor — what he calls "Local Sync" is now "Pairing Mode". Since then, we’ve shipped new improvements every day, so you might see some old features here!

<p align="center"> <a href="https://www.youtube.com/watch?v=ESZH7hd1sMY"> <img width="1312" height="738" alt="A demo of what is possible with Sculptor" src="https://img.youtube.com/vi/ESZH7hd1sMY/0.jpg" /> </a> </p>

Sculptor is built by the team at Imbue.

Download Sculptor

[!IMPORTANT] Join our Discord for dedicated support and up-to-date info from the Imbue team! Our whole team's in Discord with you, building Sculptor with Sculptor 🙂

<strong>Download Sculptor for OSX with Apple Silicon here -></strong>

<strong>Download Sculptor for Linux here -></strong>

Instructions for Apple Silicon (OSX arm64) and Linux follow. Apple Intel instructions will be updated shortly!

1. Running Sculptor

Note: You'll need an Anthropic account to use Sculptor.

1. Mac (Apple Silicon): Download the .dmg from here and double-click to unpack it into your /Applications folder. Linux: Download the .AppImage from here, mark it executable (chmod +x), and run it.

2. Enter your email and you should get to the installation wizard. To run Sculptor, you'll need to install Docker Desktop. We'll also install git and mutagen for you, if you already don't have them. As a beta tester, you'll be opted in to send error logs and basic product usage data. Let us know if this is an issue or if you need more information about this! 🙏

<p align="center"> <img width="1312" height="912" alt="Screenshot 2025-09-16 at 8 07 09 PM" src="https://github.com/user-attachments/assets/05d9a6fe-d91a-49e3-9b01-2ff31362c5e0" /> </p>

3. Provide the Anthropic credentials you'll be using. We plan on pushing out an update soon that allows Claude auth sign-in so you can use your Claude Pro and Max plans!

4. Select the repo you wish to work on. Don't worry, you can select a second repo later and switch between them freely! If there's no git repo installed there, we'll install one.

Installing Docker

On Mac: Docker Desktop

• Ensure your Docker version is 27 or higher according to docker --version
• Go to Settings > General > Virtual Machine Options after download and set “Virtual Machine Manager” to Docker VMM
• Go to Settings > Resources > Resource Allocation and configure the following:
  • Set memory limit to 8GB (or 30% of your system memory, whichever is lower)
  • Disable swap
  • Set disk limit to 50% of available disk space (minimum 100 GB)
  • Disable Resource Saver - we've seen it cause issues

On Linux: Do not install Docker Desktop.

• Instead, install Docker Engine by following the instructions here

Community

Discord

As an early tester, we highly recommend that you join our Discord community to get direct support from the Imbue team. Some interesting channels:

• #help-and-bugs: Get quick support from the Imbue team
• #show-and-tell: Share wow moments, workflows, tips, and Sculptures you're creating!
• #behind-the-scenes: Watch how the sausage gets made as our team uses Sculptor to build Sculptor

Talk with us

We’re always here for you! If you run into any issues or just want to chat with the team, book time with us!

Basic workflows

Creating agents

1. On your projects homepage, describe your task to the agent.
2. Press “Start task” or CMD+Enter to start the task. Press Enter to start a new line in your task description.
3. Your new agent will appear in the sidebar. Click into it to see the changes as the agent works.

<p align="center"> <img width="1840" height="1191" alt="Screenshot 2025-09-17 at 5 25 12 AM" src="https://github.com/user-attachments/assets/6c94daef-c100-4d4f-9e2b-27a0bff37352" /> </p>

When you’re viewing an agent’s task, you can press the + New Agent button or cmd+N to quickly create a new agent.

<p align="center"> <img width="2000" height="765" alt="sculptor_new-agent" src="https://github.com/user-attachments/assets/533f6399-a3ef-49d6-96af-5063d1219705" /> </p>

Note: for now, your first agent will take about a minute to start. We're working on making it instant!

Test the agent’s changes with Pairing Mode

Each agent runs in a safe container with a copy of your repo. Their branch lives in the container. Pairing Mode lets you bring an agent's branch into your IDE so you can instantly test its changes, while it sees your edits in real-time. When you turn it on, it checks out the agent's branch locally and keep your files + git state synced.

1. Click the Pairing Mode button in the sidebar or the top right corner to turn it on

<p align="center"> <img width="2000" height="632" alt="pairing mode" src="https://github.com/user-attachments/assets/2989e130-9995-48b6-9e50-7657c33bb0b4" /> </p>

2. Run the agent’s code, run terminal commands, make edits in your IDE, or keep working with the agent in chat while in Pairing Mode. Any local changes you make will sync back to the agent’s branch and persist after turning off Pairing Mode.

<p align="center"> <img width="2000" height="698" alt="Screenshot 2025-09-17 at 5 36 25 AM" src="https://github.com/user-attachments/assets/5bd92f8e-985e-4ac1-afb3-97d643926565" /> </p>

Note: Currently Pairing Mode will be disabled if you have uncommitted or unstashed changes in your current local branch. Stash or commit changes to enable Pairing Mode using git stash --include-untracked.

<p align="center"> <img width="1601" height="506" alt="Screenshot 2025-09-17 at 5 33 27 AM" src="https://github.com/user-attachments/assets/3d8c3ac8-ebb6-412e-b5ef-134b733403b0" /> </p>

Merging the agent’s changes

1. Turn off Pairing Mode to switch your local state back to where it was.

<p align="center"> <img src="https://github.com/user-attachments/assets/3e4799fc-a3dc-4783-bc9f-9874661b19de" alt="Image 1" width="45%" style="vertical-align: top;" /> <img src="https://github.com/user-attachments/assets/ea86c597-15d4-4965-b82c-73b89eee4869" alt="Image 2" width="45%" style="vertical-align: top;"/> </p>

2. Commit the agent’s changes in Sculptor.

<p align="center"> <img width="60%" alt="Screenshot 2025-09-17 at 5 37 45 AM" src="https://github.com/user-attachments/assets/c129b2cf-ea60-490f-8379-9fe24c281dd0" /> </p>

3. Press the Merge button in the top right corner, select your target branch, and pull the agent’s changes.

<p align="center"> <img width="60%" alt="pull agent changes" src="https://github.com/user-attachments/assets/7c30fdb1-fc29-48dd-9b45-1498f90ed386" /> </p>

For power users: you can also commit and merge the agent's changes locally while Pairing Mode is on using git commit on the branch, checking out the target branch, and git merge <source-branch>.

Resolving merge conflicts

If you expect merge conflicts, the agent can help you resolve them!

1. Push your target branch to the agent’s branch using the Merge panel.
2. Tell the agent to resolve any merge conflicts that arise.
3. After it’s done, merge the agent branch into your target branch as described in the last section (Merging the agent’s changes).

<p align="center"> <img width="60%" alt="push agent changes" src="https://github.com/user-attachments/assets/9b8cf01a-bfd0-48ff-b756-8e5cb457ea9d" /> </p> <p align="center"> <img width="2000" height="461" alt="Screenshot 2025-09-17 at 5 48 24 AM" src="https://github.com/user-attachments/assets/e4b55480-76a0-41fd-bfcb-654d0feb2dc4" /> </p>

If you try merging the agent's branch into your target branch, and merge conflicts arise, you will see this dialog. Choose whether to force merge or abort if there are conflicts that can’t be resolved.

<p align="center"> <img width="1840" height="1191" alt="Screenshot 2025-09-17 at 5 47 04 AM" src="https://github.com/user-attachments/assets/aca9db5b-782b-437b-ae5c-75d9f4268789" /> </p>

Advanced workflows

@autocomplete file and folder names

Anytime during a task, @mention a part of a filename to autocomplete it.

<p align="center"> <img width="1312" height="912" alt="Screen