One Company: An AI-Driven Development Model for Multi-Project Collaboration

Background

In modern software development, individual developers or small teams often need to maintain multiple projects simultaneously. Traditional development requires switching context between different projects, managing different tech stacks and deployment processes, leading to inefficiency and increased cognitive load.

One Company is an AI Agent-based multi-project collaboration model that achieves efficient project management and continuous delivery through unified task scheduling and specialized division of labor.

Core Concepts

1. Unified Project Registry

All projects are managed under a single "company" framework, each with:

• Independent code repository
• Dedicated tech stack
• Separate deployment pipeline
• Unified task scheduling

2. Specialized Agent Team

3. Kanban-Based Task Flow

All tasks are managed through a Kanban system, ensuring:

• Transparent task status
• Clear dependency tracking
• Progress visibility
• Quality assurance

Practical Examples

resume.blog Project

Tech Stack: Next.js 15 + React 19 + TypeScript + Tailwind CSS + SQLite

Implemented Features:

• Personal blog system
• Resume display
• Book recommendations
• Recipe sharing
• Bilingual support (Chinese/English)

Development Workflow:

1. Requirements analysis -> Jobs Product Agent creates PRD
2. Technical design -> Linus Dev Agent develops technical plan
3. Code implementation -> Linus Dev Agent writes code
4. Quality verification -> Turing QA Agent runs tests
5. Deployment -> Elon CEO Agent makes final decision

fitknow-web Project

Tech Stack: Vite + React + JavaScript

Project Purpose: Fitness knowledge frontend application based on fitness data from Bilibili creator "HaoRen SongSong"

Key Features:

• Nutrition calculator
• Training plan recommendations
• Dark mode
• PWA offline support

Technical Implementation

Multi-Project Routing

// Project registry
const projectRegistry = {
  'resume.blog': {
    repo: 'git@github.com:remixu1994/resume.blog.git',
    localPath: '~/workspace/resume.blog',
    techStack: 'Next.js',
    description: 'Personal blog, resume, portfolio'
  },
  'fitknow-web': {
    repo: 'git@github.com:remixu1994/fitknow-web.git', 
    localPath: '~/workspace/fitknow-web',
    techStack: 'Vite + JS',
    description: 'Fitness knowledge frontend app'
  }
};

Task Scheduling

# Specify target project when creating a task
task = kanban_create(
    title="feat: Add AI news category",
    assignee="linus-dev",
    body="Project: resume.blog\nRepo: remixu1994/resume.blog\nLocal path: ~/workspace/resume.blog\n\nTask: Add AI news category to blog system",
    workdir="~/workspace/resume.blog"
)

Workflow Automation

# GitHub Actions CI
name: CI Baseline
on: [push, pull_request]
jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '20'
      - run: npm ci
      - run: npm run lint
      - run: npm test
      - run: npm run build

Best Practices

1. Project Isolation

Each project is developed in an independent directory to avoid dependency conflicts:

~/workspace/
├── resume.blog/     # Next.js project
├── fitknow-web/     # Vite project
└── shared/          # Shared utilities (optional)

2. Unified Git Workflow

All projects follow the same branching and PR process:

• Branch naming: <type>/<task-id>-<description>
• PRs must pass CI checks
• Squash merge to keep history clean

3. Documentation-Driven

Every project must include:

• PROJECT.md - Project vision and scope
• ARCHITECTURE.md - Technical architecture and design decisions
• README.md - Quick start guide

4. Continuous Integration

All projects are configured with automated CI:

• Code quality checks (ESLint)
• Unit tests (Vitest)
• Build verification (Next.js/Vite build)

Future Outlook

1. Intelligent Task Assignment

Automatically assign the most suitable agent based on agent capabilities and project requirements.

2. Cross-Project Knowledge Sharing

Build a shared knowledge base so that experiences from different projects can inform each other.

3. Automated Deployment

Enable one-click deployment to different platforms (Vercel, Netlify, self-hosted, etc.).

4. AI-Assisted Code Review

Use AI for code review, providing improvement suggestions and potential issue warnings.

Summary

The One Company model solves the complexity of multi-project management through AI Agent specialization and unified scheduling. It not only improves development efficiency but also ensures code quality and project maintainability.

For individual developers or small teams, this model provides a scalable project management approach, allowing them to focus on creating value rather than being bogged down by tedious management tasks.

This article is the first in the One Company series. Future articles will dive deeper into each agent's workflow and best practices.