wiggum-cli

AI-powered feature development loop CLI - Auto-detect your tech stack and generate intelligent development environments.

 ██╗    ██╗ ██╗  ██████╗   ██████╗  ██╗   ██╗ ███╗   ███╗
 ██║    ██║ ██║ ██╔════╝  ██╔════╝  ██║   ██║ ████╗ ████║
 ██║ █╗ ██║ ██║ ██║  ███╗ ██║  ███╗ ██║   ██║ ██╔████╔██║
 ██║███╗██║ ██║ ██║   ██║ ██║   ██║ ██║   ██║ ██║╚██╔╝██║
 ╚███╔███╔╝ ██║ ╚██████╔╝ ╚██████╔╝ ╚██████╔╝ ██║ ╚═╝ ██║
  ╚══╝╚══╝  ╚═╝  ╚═════╝   ╚═════╝   ╚═════╝  ╚═╝     ╚═╝

  ██████╗ ██╗      ██╗
 ██╔════╝ ██║      ██║
 ██║      ██║      ██║
 ██║      ██║      ██║
 ╚██████╗ ███████╗ ██║
  ╚═════╝ ╚══════╝ ╚═╝

Simpson Yellow (#FED90F) branding

Installation

# Initialize in your project (recommended)
npx wiggum-cli init

# Or install globally
npm install -g wiggum-cli

Quick Start

# REPL-first: Just run wiggum to start interactive mode
wiggum

# Inside the REPL:
wiggum> /init              # Initialize (scans project, configures AI)
wiggum> /new my-feature    # Create spec with AI interview
wiggum> /run my-feature    # Run the development loop
wiggum> /exit              # Exit when done

CLI Mode (for scripts/CI)

# Traditional CLI commands still work
wiggum init                # Initialize project
wiggum new my-feature --ai # Create spec with AI
wiggum run my-feature      # Run the loop
wiggum monitor my-feature  # Monitor progress

Commands

wiggum init

Initialize Wiggum in the current project. Uses AI to analyze your codebase, detect the tech stack, and generate intelligent configuration files.

wiggum init [options]

Options:

Examples:

# Initialize with AI analysis (interactive)
wiggum init

# Initialize and enter REPL mode
wiggum init -i

# Initialize with OpenAI provider
wiggum init --provider openai

# Non-interactive mode (for CI/scripts)
wiggum init --yes

wiggum run <feature>

Run the feature development loop for a specific feature. Executes the AI-driven implementation workflow.

wiggum run <feature> [options]

Options:

Examples:

# Run the feature loop
wiggum run user-authentication

# Run with git worktree isolation
wiggum run payment-flow --worktree

# Resume an interrupted session
wiggum run payment-flow --resume

# Use specific model with iteration limits
wiggum run my-feature --model opus --max-iterations 30

wiggum monitor <feature>

Launch the monitoring dashboard to track feature development progress in real-time.

wiggum monitor <feature> [options]

Options:

Examples:

# Monitor with default settings
wiggum monitor user-authentication

# Monitor with faster refresh rate
wiggum monitor my-feature --interval 2

# Use bash monitor script
wiggum monitor my-feature --bash

wiggum new <feature>

Create a new feature specification from template or AI-powered interview.

wiggum new <feature> [options]

Options:

Examples:

# Create spec with AI interview (recommended)
wiggum new user-dashboard --ai

# Create a new spec from template
wiggum new user-dashboard

# Create and open in VS Code
wiggum new user-dashboard --edit

# Create with vim, skip confirmations
wiggum new user-dashboard --edit --editor vim --yes

# Overwrite existing spec
wiggum new user-dashboard --force

AI Mode (--ai): The AI-powered spec generation guides you through a 4-phase interview:

1. Context Gathering - Share reference URLs or files for context
2. Goals Discussion - Describe what you want to build
3. Interview - AI asks clarifying questions (3-5 questions typically)
4. Generation - AI generates a detailed, project-specific specification

Interactive REPL Mode (Default)

REPL-first: Running wiggum with no arguments opens the interactive REPL:

wiggum

This is the recommended way to use Wiggum - all commands are available interactively with persistent session state.

REPL Commands

Example Session

$ wiggum

Wiggum Interactive Mode
Not initialized. Run /init to set up this project.

wiggum> /init
# Scans project, prompts for API key, runs AI analysis...

wiggum> /new user-dashboard
# AI interview starts (no --ai flag needed in REPL)...

wiggum> /run user-dashboard
# Development loop starts...

wiggum> /exit

Benefits of REPL Mode

• Simpler mental model - One entry point, everything inside REPL
• No flags to remember - AI mode is default for /new
• Persistent session - Scan result, provider, model carry through
• Discoverable commands - /help always available

Generated Files Structure

After running wiggum init, the following structure is created:

.ralph/
├── ralph.config.cjs          # Main configuration file
├── prompts/                  # AI prompt templates
│   ├── PROMPT.md            # Implementation prompt
│   ├── PROMPT_feature.md    # Feature planning prompt
│   ├── PROMPT_e2e.md        # E2E testing prompt
│   ├── PROMPT_verify.md     # Verification prompt
│   └── PROMPT_review.md     # PR review prompt
├── scripts/                  # Automation scripts
│   ├── feature-loop.sh      # Main development loop
│   └── ralph-monitor.sh     # Monitoring script
├── specs/                    # Feature specifications
│   └── _example.md          # Example spec template
└── CLAUDE.md                # Project context for Claude

Detection Capabilities

Wiggum automatically detects and configures support for:

Core

• Frameworks: Next.js (App/Pages Router), React, Vue, Nuxt, Svelte, SvelteKit, Remix, Astro
• Package Managers: npm, yarn, pnpm, bun
• Testing: Jest, Vitest (unit), Playwright, Cypress (E2E)
• Styling: Tailwind CSS, CSS Modules, Styled Components, Emotion, Sass

Data Layer

• Databases: PostgreSQL, MySQL, SQLite, MongoDB, Redis
• ORMs: Prisma, Drizzle, TypeORM, Mongoose, Kysely
• APIs: REST, GraphQL, tRPC, OpenAPI

Frontend

• State Management: Zustand, Jotai, Redux, Pinia, Recoil, MobX, Valtio
• UI Components: shadcn/ui, Radix, Material UI, Chakra UI, Ant Design, Headless UI
• Form Handling: React Hook Form, Formik, Zod, Yup

Services

• Authentication: NextAuth.js, Clerk, Auth0, Supabase Auth, Lucia, Better Auth
• Analytics: PostHog, Mixpanel, Amplitude, Google Analytics, Plausible
• Payments: Stripe, Paddle, LemonSqueezy
• Email: Resend, SendGrid, Postmark, Mailgun

Infrastructure

• Deployment: Vercel, Netlify, Railway, Fly.io, Docker, AWS
• Monorepos: Turborepo, Nx, Lerna, pnpm workspaces

MCP (Model Context Protocol)

• Detects existing MCP server configurations
• Identifies if project is an MCP server/client
• Recommends relevant MCP servers based on stack

AI Enhancement

Wiggum uses a multi-agent AI system for deep project analysis:

wiggum init --provider anthropic

Supported Providers

Optional Services

AI Analysis Features

Wiggum uses a 4-phase multi-agent architecture:

1. Planning Orchestrator: Creates a focused analysis plan based on the detected stack
2. Parallel Workers: Context Enricher explores the codebase while Tech Researchers gather best practices (runs concurrently)
3. Synthesis: Merges worker results and detects essential MCP servers (Playwright, database)
4. Evaluator-Optimizer: QA loop that validates and improves the analysis (max 2 iterations)

Configuration

ralph.config.cjs

module.exports = {
  // Project paths
  paths: {
    root: '.ralph',
    specs: '.ralph/specs',
    prompts: '.ralph/prompts',
    scripts: '.ralph/scripts',
  },

  // Feature loop settings
  loop: {
    maxIterations: 10,
    maxE2eAttempts: 5,
    defaultModel: 'sonnet',
  },

  // Detected stack (auto-populated)
  stack: {
    framework: { name: 'Next.js', version: '14.0.0' },
    testing: {
      unit: { name: 'Vitest' },
      e2e: { name: 'Playwright' },
    },
    // ... other detected technologies
  },
};

Requirements

• Node.js 18.0.0 or higher
• Git (for worktree features)
• Claude Code CLI (for running the development loop)

Environment Variables

Contributing

Contributions are welcome! Please read our contributing guidelines before submitting a pull request.

Development Setup

# Clone the repository
git clone https://github.com/federiconeri/wiggum-cli.git
cd wiggum-cli

# Install dependencies
npm install

# Build the project
npm run build

# Run locally
node bin/ralph.js init

Running Tests

# Type checking
npx tsc --noEmit

# Run tests (when available)
npm test

License

MIT License - see LICENSE for details.

Built with Simpson Yellow (#FED90F) by the Wiggum team.