Yuva AI

A lightweight development agent framework with 12 specialized agents, 19 LLM platforms, on-demand prompts, and auto-detection.

What is this?

A development agent framework that turns your AI coding tool into a multi-agent system:

• 12 Specialized Dev Agents — Requirements, Planning, Execution, Testing, Security, Debugging, and more
• On-Demand Prompts — Agent prompts served from the package, not copied to your project
• Auto-Detection — Detects your AI tool (Claude, OpenCode, Cursor, Codex, etc.) and configures automatically
• 19 LLM Platforms — Works with Claude, GPT, Gemini, Ollama, OpenCode, Cursor, and more
• Lightweight Init — Creates only 3 files instead of 67
• Hybrid Orchestrator — CLI scans your project, AI picks the right agents
• Session Persistence — Never lose progress across conversations
• Zero Dependencies — Pure Node.js, installs in seconds

Quick Start

# Install globally
npm install -g yuva-ai

# Initialize in your project (auto-detects your AI tool)
yuva init

# Or specify your tool
yuva init opencode
yuva init cursor
yuva init codex

That's it. Open your project in your AI tool — it reads AGENTS.md and knows how to use agents.

How It Works

┌─────────────────────────────────────────────────────────────┐
│                    AGENTS.md (Orchestrator)                   │
│  AI reads this file and learns how to use the agent system   │
└─────────────────────────────────────────────────────────────┘
                              │
                    ┌─────────┴─────────┐
                    ▼                   ▼
         ┌──────────────────┐  ┌──────────────────┐
         │ yuva agent       │  │ yuva agent       │
         │ orchestrate      │  │ show <name>      │
         │                  │  │                  │
         │ Scans project:   │  │ Returns full     │
         │ - existing code? │  │ agent prompt     │
         │ - language?      │  │ from package     │
         │ - framework?     │  │ on demand        │
         │ - git status?    │  │                  │
         └──────────────────┘  └──────────────────┘
                    │                   │
                    ▼                   ▼
         ┌──────────────────────────────────────────┐
         │          DEVELOPMENT AGENTS (12)          │
         ├──────────────────────────────────────────┤
         │ Existing Code │ Requirements │ Planner   │
         │ Execution     │ Tester       │ Reviewer  │
         │ Security      │ Debugger     │ Refactor  │
         │ Continuity    │ Risk         │ State Mgr │
         └──────────────────────────────────────────┘

CLI Commands

# Setup
yuva init                  # Auto-detect AI tool and initialize
yuva init opencode         # Initialize for specific tool
yuva init cursor --force   # Reinitialize for different tool
yuva upgrade               # Migrate from old format / update
yuva doctor                # Diagnose setup issues

# Agent Commands
yuva agent show <name>     # Get full agent prompt (on demand)
yuva agent list            # List all available agents
yuva agent orchestrate     # Scan project context (JSON output)

# List & Custom Agents
yuva list                  # List all installed agents
yuva add create <name>     # Create a custom agent
yuva add remove <name>     # Remove an agent

# Multi-LLM
yuva llm list              # List supported LLMs
yuva llm use <name>        # Switch LLM platform
yuva llm detect            # Detect current LLM
yuva llm generate          # Generate configs for all LLMs

# Configuration
yuva config                # Show current config
yuva config set <k> <v>    # Set a config value

# Analytics
yuva status                # Show project status
yuva help                  # Show full help

Available Agents

Multi-LLM Support

Works with 20 AI platforms. Auto-detected on yuva init:

Commercial

Open Source / Local

Terminal / CLI

# Switch platforms anytime
yuva llm use opencode
yuva llm use cursor
yuva llm use ollama

Project Structure (After yuva init)

your-project/
├── AGENTS.md                       # Orchestrator (source of truth)
├── .aiautomations/
│   ├── config.json                 # Tool config + package path
│   └── agents.md                   # Agent index
└── .cursor/rules/yuva.mdc         # (only if Cursor detected)

Agent prompts are served on demand from the installed package — no file bloat.

Custom Agents

# Create a custom agent
yuva add create my-agent

# This creates a local override in:
# .aiautomations/prompts/my-agentagent.md

# Local agents always take priority over package agents

Upgrading from v3.2 (Old Format)

If you have the old setup with CLAUDE.md and 67+ copied files:

yuva upgrade

This will:

• Rename CLAUDE.md → AGENTS.md
• Create .aiautomations/config.json
• Create .aiautomations/agents.md index
• Remove copied prompts (now served from package)

Development

npm install
npm test
npm run test:coverage
npm run lint
npm run doctor

Works With

Commercial: Claude Code, Cursor, Windsurf, GitHub Copilot, Gemini CLI, Amazon Q, Cody, Antigravity Open Source: Ollama, LM Studio, Jan.ai, Continue.dev, Open Interpreter, LLM CLI, Tabby Terminal: OpenCode, Codex CLI, Kilo Code, Aider Models: Llama 3, Mistral, CodeLlama, DeepSeek, Qwen, Phi-3, GPT-4, Claude, Gemini

Contributing

1. Create agent prompt in template/.aiautomations/prompts/
2. Add to AGENT_MAP in lib/commands/agent.js
3. Add tests
4. Submit PR

License

MIT License - see LICENSE file for details.

Links

• NPM Package
• GitHub Repository
• Report Issues