Automaker: Build Software in Days with an Autonomous AI Studio

If you’ve ever stared at a to‑do list and wished you had a tool that could write the code for you, Automaker is the answer. It’s an open‑source, agent‑based development studio that lets you describe features on a Kanban board, then uses Claude‑powered agents to implement them almost instantly.

What Is Automaker?

Automaker is a fully autonomous AI development platform that marries a graphical interface with an underlying Claude Agent SDK. The core idea is simple:

1. Add a feature to the board – text, images, or screenshots.
2. Move the card to In Progress.
3. An AI agent takes over: it fetches the relevant files, writes code, runs tests, and commits changes in an isolated git worktree.
4. Watch the live progress via a WebSocket stream.
5. Review the diff, approve, and merge.

The result is a finished feature, ready for deployment, all driven by the same AI pipeline that powers this guide.

Key Features

Tech Stack Overview

• Frontend: React (v19), Vite (v7), TypeScript, TanStack Router, Zustand, Tailwind, Radix UI, dnd-kit.
• Backend: Node.js (ESM), Express (v5), WebSocket, node-pty for terminal.
• AI: Claude Agent SDK (anthropic‑ai) – the engine that orchestrates all agent actions.
• Development Workflow: Docker (multi‑arch amd64/arm64), GitHub CI, Vitest, Playwright, Husky.
• Shared Libraries: @automaker/* packages provide types, utilities, prompts, and infrastructure helpers.

Quick Start – From Code to Running App

# 1️⃣ Clone and install
git clone https://github.com/AutoMaker-Org/automaker.git
cd automaker
npm install

# 2️⃣ Build shared packages (optional; dev mode builds on demand)
npm run build:packages

# 3️⃣ Launch the UI
# Web app
npm run dev:web
# Or Electron desktop
npm run dev:electron

You’ll be prompted to set up authentication. The simplest way is to install the Claude Code CLI and log in – Automaker detects the credentials automatically.

Docker Deployment (Recommended for Production)

# Build images and start containers
docker compose up -d
# UI -> http://localhost:3007
# API -> http://localhost:3008

To work on a host project from inside the container, add a docker-compose.override.yml as shown in the repo to mount the desired directories and authentication keys.

Using Automaker in Practice

1. Create a Project – Open the UI, click Add Project and point to a Git repository.
2. Add Features – Write a description, attach screenshots if needed.
3. Approvals & Reviews – For spec or full planning modes, review the generated plan before execution.
4. Testing – Automated unit and E2E tests run in the background; you can also trigger a Playwright run via the API.
5. Merge – Once the agent pushes changes into a worktree, a PR is automatically created, or you can manually merge after approval.

Developers who enjoy fast prototyping will find the Kanban + agent loop invaluable. Teams can also spin up a shared Automaker instance for collaborative feature building.

Extending Automaker

Automaker is intentionally modular:

• Prompt Templates – Drop your own prompt files into packages/prompts to customize agent behavior.
• Custom Agent Profiles – Configure memory size, model (Claude Opus, Sonnet, Haiku), and plan depth.
• Hooks & Webhooks – The Express API exposes endpoints for CI/CD integration (e.g., trigger a rebuild after a PR merge).
• Plugins – The @automaker/platform package allows you to create path‑sandboxed helpers for third‑party services.

Because the codebase is written in TypeScript and uses monorepo tooling, you can fork the repo, modify apps/ui or apps/server, and re‑build with npm run dev.

Community & Support

• Discord – Join the Agentic Jumpstart community to discuss patterns, troubleshoot and share your own Automaker projects.
• GitHub Discussions – File feature requests or ask questions; the maintainers are active about pull requests.
• Documentation – Extensive docs are hosted in the repo’s docs/ directory and are also available via the web UI.

Why Automaker Matters

Traditional dev tooling writes code; Automaker directs code. This paradigm shift enables developers to focus on architecture, business logic and user experience while AI handles the heavy lifting. For startups, solo creators, or large teams, the time‑to‑product slash (TTPS) can be reduced from weeks to days.

Get Started Today

1. Clone the repo or try the live Docker image.
2. Add your Claude API key or install the Claude Code CLI.
3. Create your first feature card and watch an agent write code in real time.

Experience the future of software engineering – one AI agent at a time.