What you get
-
Zero-config project setup
Clone, launch, work. The setup agent inspects the repo and stands up everything the project needs.
Read more
-
Full Docker stack per project
Each project gets its own Docker Compose. Isolated, reproducible, and gone when you stop it.
Read more
-
Multiplayer by default
Share an agent the way you'd share a Google Doc. Anyone with the link is in the room.
Read more
-
Multiple agents per project
One project, several agents working in parallel. Each in its own session, each visible.
Read more
-
SSH into any container
ssh -p 2222 container-name@localhost. Same session the agent is in. Same shell, same prompt.
Read more
-
Works from any device
Every session is a URL. Every URL works on every device. The agent doesn't care which one is in front of you.
Read more
-
Persistent containers
A project keeps running between sessions. Come back tomorrow, the dev server is already up.
Read more
Point Loopyard at a git repo. It examines the codebase, writes a Dockerfile, spins up databases, starts the dev server, and hands you a working environment with Claude Code agents ready to write and run code inside it.
Everything is multiplayer. Multiple people can watch agents work, interact with them, and use the same terminal sessions simultaneously. Open the same project on your phone and laptop, or have three developers watching the same agent debug a failing test.
Why this exists
Running Claude Code locally is great, but it runs on your host and you manage the environment yourself. Loopyard gives each project a containerized workspace: isolated, reproducible, sharable. The agent doesn’t just write code, it builds the entire stack. Dockerfile, services, dev server, environment variables. Then it execs into that container to work.
How it works
- Launch a project. Point Loopyard at any git repo.
- Setup agent auto-detects the stack: language, framework, databases, services.
- Docker Compose builds and runs everything. Workspace container, dev server, databases.
- Agents exec into the workspace container to write code, run tests, debug issues.
- You watch, interact, and collaborate. Every view is live and multiplayer.
Compared to
Claude Code CLI
Claude Code runs on your host machine. It’s powerful but:
- No containerization. Installs and changes happen on your actual system.
- Single user. No multiplayer, no sharing sessions.
- No infrastructure management. You set up Docker, databases yourself.
- One agent at a time.
Loopyard wraps Claude Code in a containerized workspace with multiplayer. The agent gets the same capabilities, in an isolated environment it built itself.
Commander
Commander provides a GUI for managing Claude Code sessions. Loopyard goes further:
- Builds the dev environment, not just manages the session. Dockerfile, Docker Compose, services, dev server.
- Multiplayer. Commander is single-user. Loopyard lets a team watch and interact with agents at the same time.
- Container isolation. Every project gets its own Docker stack. Nothing touches your host.
- SSH access. Drop into any container from your terminal. Commander is browser-only.
- Multiple agents per project. Run setup, feature, and debug agents at once on the same codebase.
Open source
Loopyard is open source. Read the code, file issues, and send patches at github.com/loopyard/loopyard.