AG4ONE

AG4ONE - Agentic Engineering Workflow built on existing tools that streamline AI Pair Programming.

Overview

AG4ONE is a unified agentic engineering workflow that combines the best features from GWD methodology, Serena semantic analysis, and Ralph autonomous looping to create the most comprehensive AI pair programming experience available.

Primary Focus: Optimized for Opencode CLI with full compatibility for Claude Code and Gemini CLI.

🚀 What Makes AG4ONE Different

AG4ONE isn't just another AI coding tool - it's a next-generation development environment that enables even recent free models to program like flagship AI assistants:

• Three-in-One: GWD structure + Serena semantics + Ralph autonomy
• Token Efficiency: Semantic code analysis vs full file reads
• Multi-Platform: Opencode, Claude Code, Gemini CLI unified
• Intelligent Workflow: Manual → Semantic → Autonomous spectrum

Core Features

• Structured Workflow: Phase-based development with clear milestones
• Multi-Platform Compatibility: Primary support for Opencode CLI, with Claude Code and Gemini CLI compatibility
• Context Engineering: Smart management of AI context windows across different AI platforms
• Verification Systems: Built-in quality checks and testing workflows
• Documentation: Auto-generated project documentation and state tracking

Quick Start

Installation

# Install globally for all projects
npx ag4one --global

# Install locally for current project only
npx ag4one --local

# Install to custom AI CLI config directory
npx ag4one --global --config-dir ~/.opencode-custom

Platform Integration

AG4ONE automatically detects and configures for:

• Opencode CLI (primary) - 🌟 Full feature support with Serena integration
• Claude Code - 🌟 Complete workflow and skill support
• Gemini CLI - 🌟 Basic workflow support

Three Development Modes

1️⃣ Manual Workflow Mode (GWD-style)

Perfect for structured development with human oversight:

# Create and plan projects manually
Load ag4one workflow and create new project
Load ag4one workflow and create roadmap
Load ag4one workflow and execute phase

2️⃣ Semantic Analysis Mode (Serena-enhanced)

IDE-like precision for complex codebases:

# Semantic code analysis and editing
Load ag4one workflow and analyze-codebase
Load ag4one workflow and semantic-search "pattern"
Load ag4one workflow and precise-edit

3️⃣ Autonomous Loop Mode (Ralph-powered)

Hands-free autonomous development:

# Create PRD and run autonomous loop
Load ag4one PRD skill and create PRD for [feature]
./scripts/ag4one/ag4one-loop.sh [max_iterations]

Platform Command Equivalents

| Opencode CLI | Claude Code | Gemini CLI | |--------------|-------------|------------| | Load ag4one workflow and create new project | /ag4:new-project | /ag4:new-project | | Load ag4one workflow and create roadmap | /ag4:create-roadmap | /ag4:create-roadmap | | Load ag4one workflow and execute phase | /ag4:execute-phase | /ag4:execute-phase | | Load ag4one workflow and analyze-codebase | /ag4:analyze-codebase | /ag4:analyze-codebase |

Workflow Overview

AG4ONE follows a structured approach:

1. Discovery: Research and requirements gathering
2. Planning: Create detailed roadmaps with milestones
3. Execution: Build features in focused phases
4. Verification: Quality assurance and testing
5. Iteration: Continuous improvement and refinement

Available Commands

| Category | Commands | |----------|----------| | Setup | /ag4:new-project, /ag4:help, /ag4:whats-new | | Planning | /ag4:create-roadmap, /ag4:plan-phase, /ag4:discuss-phase | | Execution | /ag4:execute-phase, /ag4:execute-plan | | Verification | /ag4:verify-work, /ag4:debug | | Milestones | /ag4:new-milestone, /ag4:complete-milestone | | Session | /ag4:progress, /ag4:resume-work, /ag4:pause-work | | Utilities | /ag4:add-todo, /ag4:check-todos, /ag4:map-codebase |

Run /ag4:help for detailed command descriptions.

Architecture

AG4ONE consists of:

• Commands: Entry points for different workflow actions
• Workflows: Detailed implementation guides
• Templates: Reusable patterns and structures
• Agents: Specialized AI assistants for specific tasks
• References: Best practices and guidelines

Context Engineering

AG4ONE optimizes AI pair programming by:

• Managing context window efficiently across all supported platforms
• Providing clear, structured prompts optimized for each AI assistant
• Maintaining project state across sessions
• Using progressive disclosure for complex tasks
• Platform-specific optimizations for Opencode, Claude Code, and Gemini CLI

Quality Assurance

Built-in verification includes:

• Automated testing workflows
• Code quality checks
• User acceptance testing
• Milestone validation
• Technical debt tracking

Documentation

• AG4-STYLE.md: Development style and conventions
• CHANGELOG.md: Version history and changes
• Templates: Reusable project patterns
• Workflows: Step-by-step implementation guides

Contributing

AG4ONE follows the style guide in AG4-STYLE.md. Key principles:

• Clear, imperative language
• Structured XML tags for semantic meaning
• Progressive disclosure of information
• Context-aware workflows
• Quality-first development

License

MIT License - see LICENSE for details.

Repository

https://github.com/XcluEzy7/ag4one

AG4ONE: Streamlining AI Pair Programming through Agentic Engineering Workflows.