Your AI dev team
while you

Queue your tasks, start a batch, and walk away. Four AI agents — a lead, a developer, and two independent reviewers — work through your GitHub issues autonomously, with every PR reviewed twice before it touches main.

View on GitHub
quadwork — demo
How It Works

Queue a batch. Walk away.

Head assigns an issue to Dev. Dev opens a PR. RE1 and RE2 each review independently. Head merges and picks the next ticket. Repeat until the batch is complete.

Agent Workflow Loop
Youstart a batch
Headcreates issues + queue
Headassigns issue to Dev
Devopens a PR with code
RE1 + RE2independent reviews
Headmerges, picks next issue
repeat overnight
Batch Progress
Batch Progress (7 items)
#101
Queued
#102
Queued
#103
Queued
#104
Queued
#105
Queued
#106
Queued
#107
Queued
0/7 merged · 7 queued↻ auto-assigns next

4 Agents in One Team

Head
Team Lead
Dev
Full-stack Developer
RE1
Code Reviewer
RE2
Code Reviewer
The Four Quadrants

Four quadrants. One dashboard. Zero blind spots.

01

Four agents collaborate like a human team

The 4 agents keep chatting to work together for each task — from task assignment to implementation to two reviewers who must both approve the code changes to merge and next assignment. All automated work happens via agents chatting like humans, making decisions, arguing over code quality, and reaching consensus before any code touches main.

02

Each agent has their own desk

Using the user’s LLM subscription (Claude, Codex, Gemini) via terminal CLI — not API keys. Each agent has their own terminal and git worktree, collaborating with other agents like a human staff member has their own desk and workspace. You see exactly what each agent is doing in real-time.

03

Wake up to merged pull requests

Like a night shift team that works while you rest — QuadWork runs overnight batches using a scheduled trigger. Queue your tasks, go to sleep, and wake up to a batch of successfully merged pull requests. The progress panel tracks every issue from queued to in review to approved to merged.

04

Monitor and manage from anywhere

High-class operator features: scheduled trigger messaging to keep agents working autonomously, Telegram and Discord bridges so you can monitor and manage agent work remotely from your phone. Like having a project manager’s dashboard in your pocket.

Quick Start

Get started in 30 seconds.

terminal

That's it. The wizard handles the rest:

1.Installs AgentChattr (the chat server)
2.Creates git worktrees for 4 agents
3.Opens the dashboard at localhost:8400