DollhouseMCP 2.0

DollhouseMCP 2.0 is an open-source platform that turns prompts into modular building blocks for AI customization. It provides reusable element types including personas, skills, templates, agents, memories, and ensembles that users can create, activate, combine, version, and share. The platform forms part of Dollhouse Research, which focuses on user-owned AI customization and control.

Six element types address behavior, capabilities, structure, execution, context, and coordinated stacks. It ships with 38 bundled starter elements such as the Dollhouse expert suite and a session monitor agent. Users build elements from chat, install them from the public Dollhouse Collection, or convert compatible skills into first-class Dollhouse elements. A local portfolio stores these elements in readable files that work offline, support backup, and sync to GitHub while allowing submission of community content back to the collection.

The platform includes MCP-AQL, which groups operations into Create, Read, Update, Delete, and Execute endpoints with built-in introspection so large language models can discover server capabilities at runtime. Dynamic permissioning lets personas, skills, and ensembles act as security principals that control permitted actions, approval requirements, and outright denials. Agent execution occurs inside a controlled loop featuring a Gatekeeper, autonomy evaluator, and danger zone enforcement that provides step-by-step guidance and can hard-block high-risk operations such as destructive commands and external calls. Additional operations cover collection install and submission, portfolio sync, browser opening, enhanced search, skill conversion, execution lifecycle control, and handoff or resume flows.

Installation uses a one-command guided local web installer compatible with MCP clients including Claude Desktop, Claude Code, Cursor, VS Code, Codex, and Gemini CLI. The software is delivered as the NPM package @dollhousemcp/mcp-server and is available on GitHub. Licensing uses AGPL-3.0-or-later for the open-source core, with commercial licensing options available for proprietary, hosted, or enterprise deployments.