Tell the story. Auto writes the code.
Early Preview — We're actively battle-testing Auto with real-world clients. Expect bugs and rapid evolution. Watch and star this repo to stay updated, and join the Discord for conversations.
Building apps with AI is hit-or-miss: you prompt, get code, test it, find bugs, re-prompt, and repeat until something works (or you give up). Auto Engineer fixes this by giving AI agents deterministic scaffolds, specs, and feedback loops so they self-correct reliably.
Think of Auto like an SLR camera. In green-square mode, anyone can point and shoot; the system handles the complexity automatically. Switch to manual, and you control every parameter. Same tool, different depths. Beginners ship apps on day one; experts fine-tune every stage of the pipeline.
You model your apps using Scenes, a flow-of-time DSL where you tell the story of your application moment by moment, like a user journey. The pipeline transforms these high-level flow models into production-ready code: scenes become a domain model, which scaffolds a backend; an AI architect generates a user experience architecture, which scaffolds a frontend. Both are then implemented and tested by AI agents with deterministic feedback loops.
Auto Engineer is for teams who want to collaborate with non-technical stakeholders on real specifications, not mock wireframes, while keeping full control over the generated architecture through customizable pipelines.
npx create-auto-app@latest my-project
cd my-project
cp .env.template .env # Add your API key (Anthropic recommended)
autoYou should see server running on http://localhost:5555. Open the URL and click through to your sandbox to see the visual counterpart of your scenes.
Next steps:
- Walk through the minimal example (~5 min, single-plugin pipeline)
- Walk through the typical example (~10 min, full-stack app)
- Join the Discord community
flowchart LR
A[Scenes] --> B[Domain Model]
B --> C[Server Scaffold]
B --> D[UX Architecture]
D --> E[Frontend Scaffold]
C --> F[AI Implementation]
E --> F
F --> G[Quality Checks]
G -->|Fail| F
G -->|Pass| H[Production Code]
Scenes define your application as moments of behavior. The pipeline converts these to a domain model, scaffolds both server and frontend code with implementation hints, then AI agents implement the code. If tests fail, the AI receives error feedback and self-corrects. Passing code undergoes type checking, linting, and runtime validation.
| Package | Description |
|---|---|
@auto-engineer/cli |
Command-line interface for running Auto Engineer pipelines |
@auto-engineer/pipeline |
Command/event pipeline orchestration with projections and reactors |
@auto-engineer/message-bus |
In-process message bus for command dispatch and event publishing |
@auto-engineer/message-store |
Event persistence and replay with pluggable backends |
@auto-engineer/narrative |
DSL for modeling application behavior as scene-based flows |
@auto-engineer/job-graph-processor |
DAG job processor with dependency tracking and parallel dispatch |
@auto-engineer/id |
Deterministic ID generation for pipeline correlation |
@auto-engineer/model-factory |
Domain model generation from narratives |
| Package | Description |
|---|---|
@auto-engineer/server-generator-apollo-emmett |
Apollo GraphQL + Emmett event-sourced server scaffolding |
@auto-engineer/server-generator-nestjs |
NestJS server scaffolding |
@auto-engineer/generate-react-client |
React + Vite + Tailwind CSS v4 + shadcn client scaffolding |
@auto-engineer/react-gen |
React code generation utilities |
create-auto-app |
Project scaffolding CLI with templates |
| Package | Description |
|---|---|
@auto-engineer/server-implementer |
AI-powered server code implementation |
@auto-engineer/app-implementer |
AI-powered full application implementation |
@auto-engineer/app-assembler |
Application assembly from generated components |
@auto-engineer/component-implementor-react |
AI-powered React component implementation |
@auto-engineer/component-parser |
Component definition parsing and analysis |
@auto-engineer/set-theme |
Theme configuration for generated applications |
| Package | Description |
|---|---|
@auto-engineer/dev-server |
Development server with SSE events and pipeline visualization |
@auto-engineer/checks |
Code validation (types, lint, tests) |
@auto-engineer/file-store |
File system operations with caching |
@auto-engineer/file-upload |
File upload handling |
@auto/surgical-edit-engine |
Precise code patching and diff application |
@auto-engineer/adk-claude-code-bridge |
Bridge between ADK and Claude Code |
@auto-engineer/submit-bug-report |
Automated bug report submission |
@auto-engineer/release-automation |
Automated release management with changesets |
| Example | Description | Complexity |
|---|---|---|
minimal |
Single-plugin pipeline with type checking | Beginner |
typical |
Full-stack task management app with event sourcing and React | Intermediate |
support-files |
Shared design system assets and tokens | Reference |
- Node.js 20.0.0+
- pnpm 10.15.0+
- AI Provider API Key (Anthropic recommended)
git clone https://github.com/BeOnAuto/auto-engineer.git
cd auto-engineer
pnpm install
pnpm watch| Command | Description |
|---|---|
pnpm watch |
Build all packages in watch mode |
pnpm build |
Build all packages |
pnpm test |
Run all tests |
pnpm check |
Install, build, lint, type-check, and test |
pnpm clean |
Clean build artifacts |
pnpm clean:check |
Clean then full check |
Contributions welcome! Please open an issue or pull request on GitHub.
Elastic License 2.0 (EL2) — See LICENSE for details.