Skip to content

Codingbuddy orchestrates 29 specialized AI agents to deliver code quality comparable to a team of human experts through a PLAN → ACT → EVAL workflow.

License

Notifications You must be signed in to change notification settings

JeremyDev87/codingbuddy

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

463 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

English | 한국어 | 中文 | 日本語 | Español

Codingbuddy

CI npm version License: MIT

Codingbuddy Multi-Agent Architecture

AI Expert Team for Your Code

Codingbuddy orchestrates 35 AI agents to deliver human-expert-team-level code quality.

A single AI can't be an expert at everything. Codingbuddy creates an AI development team—architects, developers, security specialists, accessibility experts, and more—that collaborate to review, verify, and refine your code until it meets professional standards.


The Vision

The Problem

When you ask an AI to write code, you get a single perspective. No security review. No accessibility check. No architecture validation. Just one AI doing everything "okay" but nothing excellently.

Human development teams have specialists:

  • Architects who design systems
  • Security engineers who find vulnerabilities
  • QA specialists who catch edge cases
  • Performance experts who optimize bottlenecks

Our Solution

Codingbuddy brings the specialist team model to AI coding.

Instead of one AI trying to do everything, Codingbuddy coordinates multiple specialized agents that collaborate:

┌─────────────────────────────────────────────────────────────┐
│                    Your Request                              │
│            "Implement user authentication"                   │
└─────────────────────────────────────────────────────────────┘
                            │
                            ▼
┌─────────────────────────────────────────────────────────────┐
│ 📋 PLAN: Solution Architect + Architecture Specialist       │
│          → Design system architecture                       │
│          → Define security requirements                     │
└─────────────────────────────────────────────────────────────┘
                            │
                            ▼
┌─────────────────────────────────────────────────────────────┐
│ 🚀 ACT: Backend Developer + Test Strategy Specialist        │
│         → Implement with TDD                                │
│         → Follow quality standards                          │
└─────────────────────────────────────────────────────────────┘
                            │
                            ▼
┌─────────────────────────────────────────────────────────────┐
│ 🔍 EVAL: Code Reviewer + Parallel Specialists               │
│          🔒 Security    → JWT vulnerabilities?              │
│          ♿ Accessibility → WCAG compliance?                 │
│          ⚡ Performance  → Optimization needed?              │
│          📏 Quality      → SOLID principles?                │
└─────────────────────────────────────────────────────────────┘
                            │
              ┌─────────────┴─────────────┐
              │                           │
        Critical > 0?              Critical = 0 AND
        High > 0?                  High = 0
              │                           │
              ▼                           ▼
        Return to PLAN              ✅ Quality Achieved
        with improvements           Ship with confidence

Multi-Agent Architecture

3-Tier Agent System

Tier Agents Role
Mode Agents (4) plan-mode, act-mode, eval-mode, auto-mode Workflow orchestration
Primary Agents (16) solution-architect, technical-planner, frontend-developer, backend-developer, +12 more Core implementation
Specialist Agents (15) security, accessibility, performance, test-strategy, +11 more Domain expertise

Agent Collaboration Example

When you request a feature, agents automatically collaborate:

🤖 solution-architect    → Designs the approach
   └── 👤 architecture-specialist  → Validates layer boundaries
   └── 👤 test-strategy-specialist → Plans test coverage

🤖 backend-developer     → Implements the code
   └── 👤 security-specialist      → Reviews auth patterns
   └── 👤 event-architecture       → Designs message flows

🤖 code-reviewer         → Evaluates quality
   └── 👤 4 specialists in parallel → Multi-dimensional review

Quality Assurance Cycle

The PLAN → ACT → EVAL Loop

Codingbuddy enforces a quality-driven development cycle:

  1. PLAN: Design before coding (architecture, test strategy)
  2. ACT: Implement with TDD and quality standards
  3. EVAL: Multi-specialist review (security, performance, accessibility, quality)
  4. Iterate: Continue until quality targets met

AUTO Mode: Autonomous Quality Achievement

# Just describe what you want
AUTO: Implement JWT authentication with refresh tokens

# Codingbuddy automatically:
# → Plans the implementation
# → Writes code following TDD
# → Reviews with 4+ specialists
# → Iterates until: Critical=0 AND High=0
# → Delivers production-ready code

Exit Criteria

Severity Must Fix Before Ship
🔴 Critical Yes - Immediate security/data issues
🟠 High Yes - Significant problems
🟡 Medium Optional - Technical debt
🟢 Low Optional - Enhancement

What Makes It Different

Traditional AI Coding Codingbuddy
Single AI perspective 35 specialist agent perspectives
"Generate and hope" Plan → Implement → Verify
No quality gates Critical=0, High=0 required
Manual review needed Automated multi-dimensional review
Inconsistent quality Iterative refinement until standards met

Quick Start

Prerequisites

  • Node.js 18.x or higher
  • npm 9.x+ or yarn 4.x+
  • A supported AI tool (Claude Code, Cursor, GitHub Copilot, etc.)

Installation

# Initialize your project
npx codingbuddy init

# Add to Claude Desktop config
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "codingbuddy": {
      "command": "npx",
      "args": ["codingbuddy", "mcp"]
    }
  }
}

Start Using

PLAN: Implement user registration with email verification
→ AI team plans the architecture

ACT
→ AI team implements with TDD

EVAL
→ AI team reviews from 8+ perspectives

AUTO: Build a complete auth system
→ AI team iterates until quality achieved

Full Getting Started Guide →

Claude Code Plugin (Optional)

For enhanced integration with Claude Code:

# Add the marketplace
claude marketplace add JeremyDev87/codingbuddy

# Install the plugin
claude plugin install codingbuddy@jeremydev87

# Install MCP server for full functionality
npm install -g codingbuddy
Documentation Description
Plugin Setup Guide Installation and configuration
Quick Reference Commands and modes at a glance
Architecture How plugin and MCP work together

Terminal Dashboard (TUI)

Codingbuddy includes a built-in terminal UI that displays real-time agent activity, task progress, and workflow state alongside your AI assistant.

Quick Start

# Launch MCP server with TUI auto-enabled
npx codingbuddy mcp --tui

Features

Panel Description
FlowMap Visual pipeline showing active agents, stages, and progress
FocusedAgent Live view of the currently active agent with sparkline activity
Checklist Task completion tracking from PLAN/ACT/EVAL context
Activity Chart Real-time tool invocation bar chart
Multi-session Multiple Claude Code sessions share a single TUI window

Supported AI Tools

Tool Status
Claude Code ✅ Full MCP + Plugin
Cursor ✅ Supported
GitHub Copilot ✅ Supported
Antigravity ✅ Supported
Amazon Q ✅ Supported
Kiro ✅ Supported
OpenCode ✅ Supported

Setup Guides →


Configuration

AI Model Settings

Configure the default AI model in codingbuddy.config.json:

{
  "ai": {
    "defaultModel": "claude-sonnet-4-20250514"
  }
}
Model Best For
claude-opus-4-* Complex architecture, deep analysis
claude-sonnet-4-* General development (default)
claude-haiku-3-5-* Quick lookups (not recommended for coding)

Verbosity Settings

Optimize token usage with verbosity levels:

{
  "verbosity": "compact"
}
Level Use Case
minimal Maximum token savings, essential info only
compact Balanced, reduced formatting (default)
standard Full formatting, structured responses
detailed Extended explanations, examples included

Documentation

Document Description
Getting Started Installation and quick setup
Philosophy Vision and design principles
Agent System Complete agent reference
Skills Library Reusable workflow skills (TDD, debugging, PR, etc.)
Supported Tools AI tool integration guides
Configuration Config file options
API Reference MCP server capabilities

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

License

MIT © Codingbuddy

About

Codingbuddy orchestrates 29 specialized AI agents to deliver code quality comparable to a team of human experts through a PLAN → ACT → EVAL workflow.

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Sponsor this project

 

Contributors

Languages