uhlala.app

uhlala.app

multi-agent AI development

A framework for running parallel AI agents that build production software — coordinated through GitHub, governed by four files, shipping real code.

Concept

Instead of one developer prompting one AI, uhlala runs a team of specialized agents in parallel. Each agent owns a domain, works in isolated branches, and ships through pull requests — just like a human dev team.

Principles

  • GitHub as source of truth — all work flows through issues, branches, and PRs
  • 4 governance files — AGENTS.md, PLAN.md, CONVENTIONS.md, CLAUDE.md define the system
  • 1 task = 1 agent session — each GitHub issue becomes an isolated agent run
  • Human-in-the-loop — agents propose, humans approve via PR review

Agent Team

Dan
DevOps & infrastructure
Pia
Triage & project management
Nora
Testing & quality assurance
Quinn
Backend & API development
Frida
Frontend & UI

Live Demo

It all starts with good user stories.

-> User Stories — describe a feature, get a structured story, publish straight to GitHub
built by humans and agents