devAIteam 🚀

The most advanced autonomous, local AI software development team platform

devAIteam is a state-of-the-art multi-agent platform built on top of LangGraph and optimized to run 100% locally with hybrid-thinking language models (such as Qwen3.6-35B via MLX or Ollama). The platform orchestrates an entire virtual software development team that collaborates sequentially and in parallel to transform a simple user requirement into real, fully-tested, human-reviewed, production-ready code.

---

🏛️ Multi-Agent Pipeline Architecture

The virtual team's workflow follows the traditional Software Development Life Cycle (SDLC) in an automated, highly-structured manner:

graph TD
  User([User / CLI]) -->|Requirement| PM[📋 PM Agent]
  PM -->|User Stories / US| Arch[🏗️ Architect Agent]
  Arch -->|Clean Architecture / DB / API| UI[🎨 UI Designer Agent]
  UI -->|Stitch Mockups & CSS| Dev[💻 Developer Agents]
  
  subgraph Parallel Dev
    Dev -->|Backend Code NestJS| DevBE[⚙️ Backend Dev]
    Dev -->|Frontend Code Flutter| DevFE[📱 Frontend Dev]
  end
  
  DevBE --> QA[🧪 QA Agent]
  DevFE --> QA
  
  subgraph Parallel QA & Code Quality
    QA -->|Jest/Flutter Spec Tests| TestGen[🧪 Test Generator]
    QA -->|Vulnerability & Bugs Review| CodeCheck[🔍 Code Auditor]
  end
  
  QA -->|Test & Score 0-100| Rev[👁️ Reviewer Agent]
  
  subgraph Human-in-the-Loop Interruption
    Rev -->|Code Review Round| HITL{Approve changes?}
    HITL -->|Reject and Feed Back| Rev
  end
  
  HITL -->|Approve and Finalize| DevOps[🚀 DevOps Agent]
  DevOps -->|Docker Compose / Local Preview| Registry[(💾 Central Registry)]
  Registry -->|deploy| Deploy[☁️ Deploy Agent]

Loading

---

🌟 Key Features

1. Complete Virtual Team: 8 specialized agent roles collaborating seamlessly to build your application.
2. Real Human-in-the-Loop: The Code Reviewer conducts a meticulous code audit, proposes exact patch snippets, and pauses execution waiting for human approval or feedback to trigger refinement rounds.
3. Smart Local Preview: The DevOps Agent automatically provisions a local stack with docker-compose (NestJS + Nginx + PostgreSQL) or runs local background processes (Node + Flutter), falling back gracefully to LLM-generated instructions if tools are missing.
4. Powerful Multi-command CLI: Control the entire suite using the native devAIteam executable command to generate, list, deploy, or clean up projects.
5. Centralized Project Registry: Real-time tracking of code file counts, QA quality scores, GitHub Pull Request URLs, local folder sizes, and cloud deploy status under ./output/.registry.json.
6. Cloud Deployment Integration: Direct deployments of your projects to Fly.io, Railway, and Render based on automatically detected local tools.
7. Safe Cleanups (Soft & Total Modes): Remove local code folders to free up disk space, or perform a total deletion that connects with the GitHub MCP to close PRs, delete heads/branches, and clean up repos automatically.

---

🛠️ Virtual Agent Roles

• 📋 Product Manager (PM Agent): Analyzes user requirements and drafts detailed User Stories with priority levels and story point estimations.
• 🏗️ Software Architect (Architect Agent): Designs the project topology (Clean Architecture), API REST endpoints, Entity-Relationship database models, and Mermaid diagrams with Context7 MCP support.
• 🎨 UI/UX Designer (Designer Agent): Formulates design systems and interactive screens utilizing the Google Stitch API.
• 💻 Developers (Backend & Frontend Agents): Write parallel, complete, and functional NestJS (TypeScript) and Flutter (Dart) source files with no placeholders, powered by the Filesystem MCP.
• 🧪 QA Agent (Test & Quality Auditor): Audits security/reliability (presents a 0 to 100 quality score) and autogenerates robust unit and widget test files.
• 👁️ Tech Lead / Code Reviewer (Reviewer Agent): Reviews the source code and QA outputs, manages human approvals, and commits refinement parches via the GitHub MCP.
• 🚀 DevOps Engineer (DevOps Agent): Manages local Docker environments, orchestrates browser previews, and manages startup verification.
• ☁️ Cloud Deployer (Deploy Agent): Packages deployment configurations and scripts for instant cloud launch.

---

📦 Installation and Setup

Prerequisites

• Python 3.12 or higher.
• uv (An ultra-fast Python package installer).
• Local MLX or Ollama server running at http://localhost:8000/v1 with the model Qwen3.6-35B-A3B-UD-MLX-4bit active.

Step 1: Clone and Install Dependencies

git clone https://github.com/nicolasmg-pr/devAIagent.git
cd devAIagent
uv sync

Step 2: Configure Environment Variables

Create or modify your .env file in the project root to enable credentials for your virtual agents (this file is securely protected in .gitignore):

# Cloud Deploy Tokens (Optional)
FLY_API_TOKEN=your_fly_token_here
RAILWAY_TOKEN=your_railway_token_here
RENDER_API_KEY=your_render_api_key_here

# GitHub Integrations (Highly Recommended for Code Review & RM commands)
GITHUB_OWNER=your_github_username_here
GITHUB_REPO=your_github_repo_name_here
GITHUB_PERSONAL_ACCESS_TOKEN=your_github_pat_token

# Google Stitch Keys (Optional for UI Designer)
STITCH_API_KEY=your_stitch_api_key_here

Step 3: Activate and Use the CLI

Activate your virtual environment:

source .venv/bin/activate

The devAIteam command is now fully registered in your shell!

---

🎮 CLI Usage Guide

1. Initiate a New Project

To start the pipeline and generate an application from scratch:

devAIteam "I want a mobile app to manage cooking recipes and weekly nutritional menus"

If you run devAIteam with no arguments, a gorgeous Unicode dashboard will load and prompt you for the project description interactively.

2. List Generated Projects

Get a beautifully formatted table representing all generated projects, scores, folder sizes, and links:

devAIteam list

3. Deploy to the Cloud

Deploy your selected local project to Fly.io, Railway, or Render instantly:

devAIteam deploy expensemaster-mobile

4. Delete Project (Soft Mode)

Safely deletes the local directory ./output/{project} to free up disk space, keeping remote repositories and deployments intact:

devAIteam rm expensemaster-mobile

5. Deletion of All Resources (Total Mode)

Stops containers, purges the local folder, cleans the index in .registry.json, and triggers the GitHub MCP to close pull requests and delete repository branches automatically:

devAIteam rm expensemaster-mobile --all