LoopCodeLab, our multi-agent platform for building software
LoopCodeLab is the multi-agent development platform we build at ProAIG. You describe the result you want and review a plan of stories with acceptance criteria. Several AI coding agents, such as Claude Code, Codex and Qwen, then build in parallel while a master agent reviews and integrates their work. It runs on our own servers.
What does LoopCodeLab do?
You go from an idea to a checked result, approving the plan along the way.
Describe the result you want: who it is for, the problem it solves and what done should look like.
Review the plan: LoopCodeLab turns the idea into editable stories with clear acceptance criteria.
Watch the team build: agents work on the stories while a master agent reviews and integrates their work.
Open the result: depending on what you built, you get a live preview, a repository, an app package or Drive files.
With LoopCodeLab you can build web apps, Flutter mobile apps, social video, image sets, music or voiceover, documents, spreadsheets, slides and Windows packages.
What does a build look like?
Real screens from the product; the recording of a full run follows.
The Watch it build panel in progress 
The plan's four steps, in the docs 
The LoopCodeLab home page on mobile
20 to 40 second video in use: [[TODO: CG-014]]
How is it built?
A planner agent turns your idea into stories. Each worker agent builds one or more stories in its own Git worktree, in its own tmux pane. A master agent reviews, accepts and integrates everyone's work before the final check.
- PlannerTurns your idea into stories with acceptance criteria.
- WorkersEach one builds one or more stories, in its own Git worktree and tmux pane.
- Master agentReviews, accepts and integrates every worker's work.
- Final checkThe result passes its checks before it ships.
It works with Claude Code, Codex and Qwen, and other providers. You connect your own AI keys or sign in with your subscription.
Where does it run?
At ProAIG we run it on our own servers. Its public docs also say it can run on your own server with your own AI accounts. There is a free tier to sign up and start building.
What happened at Talent Arena 2026?
Top 5 at Talent Arena 2026 Barcelona: we achieved it with the CitaSegura project, in the GSMA Open Gateway Hackathon. Proof link: [[TODO: CG-005]]
Get LoopCodeLab updates
Sign-up on loopcodelab.com is open today, so you can start building right away. If you would rather we let you know about product updates, leave your email with your consent.
Frequently asked questions
What is a multi-agent platform?
It is a system where several AI agents work at the same time on different parts of one project. A master agent reviews and integrates their work as they go. In LoopCodeLab, each worker agent builds in its own Git worktree and tmux pane.
What AI agents can I use?
LoopCodeLab works with Claude Code, Codex and Qwen, and other providers. You connect your own API keys or sign in with each provider's subscription.
Do I need my own API keys?
Not always. You can start with the free plan's included models, or connect your own keys and subscriptions for the agents you use. With your own keys, nobody on our side meters or bills your model use.
Can I use it on my own server?
According to the public docs, LoopCodeLab can run on your own server with your own AI accounts. At ProAIG we run it on our servers.
What can I build with LoopCodeLab?
Web apps, Flutter mobile apps, social video, image sets, music or voiceover, documents, spreadsheets, slides and Windows packages, according to the public docs.
Want to talk about your case?
In 20 minutes we look at your operations together and see whether AI fits.
Online booking will appear in this section.
Book your 20-minute AI readiness call