Three Minds 🧠🧠🧠

三个臭皮匠顶个诸葛亮 - A Multi-Agent Collaboration System

Three AI agents with different personas working together on the same codebase. Not just talking—they actually read files, write code, and run tests.

Features

• 🔧 Real Execution - Each agent can read files, write code, run tests via Claude Code CLI
• 👥 Multi-Perspective - Three agents with different expertise review each other's work
• ✅ Consensus Voting - All must vote YES to finish, ensuring quality
• 📁 Shared Workspace - Collaborate on the same project directory
• 📝 Full Transcript - Auto-saves discussion and changes history

Installation

# Clone
git clone https://github.com/Enderfga/three-minds.git
cd three-minds

# Install dependencies
npm install

# Build
npm run build

# Link globally (optional)
npm link

Requirements

• Node.js 18+
• Claude Code CLI (claude command must be available)

Usage

# Basic usage
three-minds "Review and improve this project's code quality" --dir ./my-project

# Use code-review preset (security + performance + quality trio)
three-minds "Review all code in src/" --config code-review --dir ./project

# Specify max rounds
three-minds "Refactor this module" --dir ./module --max-rounds 5

# Save result to JSON
three-minds "task description" --dir ./project --output result.json

Preset Configurations

Default - Code Collaboration Trio

• 🏗️ Architect - Code structure, design patterns, scalability
• ⚙️ Engineer - Code quality, error handling, performance
• 🔍 Reviewer - Code standards, potential bugs, documentation

code-review - Code Review Trio

• 🛡️ Security Expert - Vulnerabilities, injection risks, permissions
• ⚡ Performance Engineer - Algorithm complexity, memory, query optimization
• ✅ Quality Reviewer - Readability, naming conventions, test coverage

idea-brainstorm - Research Brainstorm Trio

• 📚 Literature Expert - Related work, theoretical foundation
• 💡 Creative Thinker - Novel approaches, unconventional ideas
• 🔬 Feasibility Analyst - Technical constraints, implementation path

paper-writing - Paper Writing Trio

• 📝 Content Reviewer - Argument structure, logical flow
• ✍️ Language Editor - Grammar, clarity, academic tone
• 🎨 Presentation Advisor - Figures, tables, visual organization

Custom Configuration

Create a JSON config file:

{
  "name": "My Custom Trio",
  "agents": [
    {
      "name": "Expert A",
      "emoji": "🎯",
      "persona": "You are a... focusing on..."
    },
    {
      "name": "Expert B",
      "emoji": "🔬",
      "persona": "You are a... specializing in..."
    },
    {
      "name": "Expert C",
      "emoji": "📊",
      "persona": "You are a... responsible for..."
    }
  ],
  "maxRounds": 10,
  "projectDir": "."
}

Then: three-minds "task" --config ./my-config.json

Workflow

┌──────────────────────────────────────────┐
│              Round N                      │
├──────────────────────────────────────────┤
│  🏗️ Architect                            │
│  → Read files, review structure          │
│  → Execute necessary refactoring         │
│  → Vote [CONSENSUS: YES/NO]              │
├──────────────────────────────────────────┤
│  ⚙️ Engineer                             │
│  → Review architect's changes            │
│  → Add implementation details, fix bugs  │
│  → Vote [CONSENSUS: YES/NO]              │
├──────────────────────────────────────────┤
│  🔍 Reviewer                             │
│  → Review all changes                    │
│  → Check standards, bugs, docs           │
│  → Vote [CONSENSUS: YES/NO]              │
└──────────────────────────────────────────┘
          ↓
    All YES? → Done
          ↓ NO
    Continue to next round...

Output

1. Terminal Output - Real-time progress and votes from each agent
2. Markdown Transcript - Auto-saved to three-minds-{timestamp}.md in working directory
3. JSON Result - Use --output to save complete session data

Use Cases

• Code Review - Multi-angle review of PRs or code changes
• Refactoring - Collaborative complex code refactoring
• Feature Development - From design to implementation
• Bug Fixing - Locate issues and verify fixes
• Documentation - Improve and complete project docs
• Research Brainstorming - Evaluate research ideas from multiple angles
• Paper Writing - Review and improve academic papers

Notes

• Each agent will actually modify files—recommend using on a git branch
• Default max 15 rounds, adjustable via --max-rounds
• If consensus can't be reached, check if task description is clear

License

MIT