mcp-server

Overview

DollhouseMCP is a modular MCP server that allows users to create, edit, and activate various AI customization elements including personas, skills, templates, agents, memories, and ensembles. It features a portfolio-based system where users can store their custom elements and browse a community collection. The server implements a unique MCP-AQL approach with 5 semantic endpoints (Create, Read, Update, Delete, Execute) that provide clearer semantics and better permission control than traditional flat tool lists.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• mcp_aql_create

  Create new elements or install from collection

  Try:Create a new skill for data analysisTry:Install the cybersecurity expert persona from the collection
• mcp_aql_read

  List, search, get details, activate, and introspect elements

  Try:List all available personasTry:Search for skills related to code reviewTry:Activate the debug detective persona
• mcp_aql_update

  Edit existing elements

  Try:Update the code review persona to add security checksTry:Modify the research agent's execution parameters
• mcp_aql_delete

  Remove elements or clear entries

  Try:Delete the old draft personaTry:Clear all memories from last week
• mcp_aql_execute

  Run agents, manage execution lifecycle, and confirm operations

  Try:Run the research agent to find market trendsTry:Execute the bug fix agent on this codebase
• convert_skill_format

  Convert between agent skills and Dollhouse Skills format

  Try:Convert this agent skill to Dollhouse formatTry:Export the translation skill as an agent skill
• introspect

  Discover operations, parameters, element formats, and usage examples

  Try:What operations are available?Try:Show me the format for creating a persona

Comparable tools

Installation

npx @dollhousemcp/mcp-server@latest --web

claude mcp add -s user dollhousemcp -- npx -y @dollhousemcp/mcp-server

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

FAQ

What platforms does DollhouseMCP support?

It supports Claude Code, Claude Desktop, Cursor, Gemini, Codex, and local LLMs. Core element management works across all platforms, with advanced features tested extensively on Claude Code.

How do Dollhouse elements differ from regular AI prompts?

Dollhouse elements change not just the AI's behavior but also what tools it can access, commands it can run, and operations requiring approval. They act as security principals with permission policies.

Compare mcp-server with