DollhouseMCP - Vibe Tooling for AI

Your Personal AI Toolkit

Every tool you create becomes part of your permanent collection.
Customize AI to match how you actually work.
From simple prompts to complex workflows—build the AI tools you need.

💬

Create Your Tools with Natural Language

Just describe what you need. If it doesn't work just right, modify it. Make a new version. Your tools constantly improve.

🧩

Mix and Match Like Lego Blocks

Combine personas, skills, and other elements you create with tools others have shared. Snap them together in different ways to help AI do exactly what you want, predictably.

🚀

Hundreds of MCP-Compatible Platforms

Works with Claude Code and Desktop and any MCP-compatible system. Create once, use everywhere.

💾

Save What Works

Every prompt you perfect, every persona you create, every workflow you build, all become part of your permanent collection.

🌟

Share What's Great

Share your best tools with others. Discover what's already been built. Learn from real implementations that solve real problems.

📈

Constantly Improve

Your tools evolve with you. Refine based on what works. Version control keeps track as you make things better.

Getting Started

# Option 1: Local installation (recommended)
mkdir ~/mcp-servers && cd ~/mcp-servers
npm init -y
npm install @dollhousemcp/mcp-server

# Option 2: Use npx (no installation needed)
Run directly without installing:
npx @dollhousemcp/mcp-server

# Configure Claude Desktop
Add to your config file:
• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

For local installation:
{
  "mcpServers": {
    "dollhousemcp": {
      "command": "node",
      "args": ["/Users/YOU/mcp-servers/node_modules/@dollhousemcp/mcp-server/dist/index.js"]
    }
  }
}

For npx:
{
  "mcpServers": {
    "dollhousemcp": {
      "command": "npx",
      "args": ["@dollhousemcp/mcp-server"]
    }
  }
}

# Restart Claude Desktop and you're ready!
Your AI now has access to persistent tools that remember and evolve.

The Story Behind DollhouseMCP

Hey everyone,

It's been a while since I've posted about my work.

Over two years, I've been building with AI tools, searching for something with real spark. I've worked on Chrome extensions, robotics demos, writing tools, collaborative platforms. Some were interesting, some optimized my workflow, but nothing felt like it could both help people and be a product people would actually love... until recently.

I've been working with ChatGPT, Claude, Claude Code, Gemini, Groq, and others. I found cool and useful prompts scattered across Reddit, Discord, and LinkedIn, but it became messy and hard to find what I wanted when I needed it. I had an incredibly bad system for organizing them.

So I started building a catalog system for prompts using Claude Code and MCP. Day one: proof of concept worked. Week two: hey, this is pretty cool. Month two: I can create a collection for everyone to use and share. I just finished my second month and it all works remarkably well.

I call it DollhouseMCP. The simplest way to describe it is "vibe tooling" instead of vibe coding.

Vibe coding's ideal is zero to 100% done in one prompt. When that fails (always), you tweak and try again. Maybe you get something good. Even when it works, you can't repeat that success reliably. You haven't built tools to make the next project easier.

With DollhouseMCP and vibe tooling, you build individual tools using natural language. These tools persist. They can be modified, even evolve on their own as needed. The system saves everything. When multiple personas and skills interact, the whole becomes greater than the sum of its parts.

Tell it once "be a harsh security-focused code reviewer," and it remembers forever. Need it less harsh? Say "dial back the harshness and add Web security focus." Version 1.2, done.

The personas have personality and skills: analyzing LinkedIn profiles, copy editing, generating audio summaries that speak in clear, contextual language. The audio skill took five minutes to build, which felt incredibly sci-fi. They can maintain context across sessions and platforms.

For those who remember Tomorrowish, this builds on more than a decade of ML/AI work. I think the patents in natural language processing we received provided me a unique lens for blending programmatic tools with LLMs.

It's open source. It's on GitHub. It works today.

This platform is incredibly useful every day. I'm using it right now for copy editing through verbal instructions. It's helped me solve dozens of problems. And it's remarkably fun to use.

I'll continue developing this as open source. The core will always be free, with compelling business applications worth exploring. People are already sharing creative uses I never dreamed of.

For my friends in media tech: there are exciting possibilities here beyond typical AI applications.

There's a lot more to explore here.

Start Building Your AI Toolkit

Create tools that work the way you need them to.
Save what works. Share what's great.

View on GitHub Quick Start Guide