# DollhouseMCP/mcp-server [Health: Active]

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/DollhouseMCP/mcp-server  
**GitHub Stars:** 46  
**npm Downloads (last month):** 1457  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/dollhousemcp-mcp-server

## Description
One-line installable MCP server that adds reusable customization elements — personas, skills, templates, agents, memory, and ensembles (collected customization tools) — to any MCP Client application. Dynamic permissioning for safe AI operations, a robust validation architecture, versioning, and a public collection of shareable content. Install: npx @dollhousemcp/mcp-server@latest --web.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

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

## Documentation

## What DollhouseMCP/mcp-server does

DollhouseMCP/mcp-server MCP server provides modular customization elements for MCP-compatible AI applications. Its six element types are personas, skills, templates, agents, memories, and ensembles. Personas influence behavior, tone, expertise, priorities, and permission policies. Skills add on-demand capabilities, templates standardize generated outputs, agents manage multi-step goals, memories retain structured context, and ensembles combine multiple elements into one activatable unit.

The server is useful for teams or individual developers who want reusable AI behavior instead of repeating the same instructions in every conversation. It also supports a public collection of shareable elements and local installations from that collection.

## How it works

Elements are kept in a local portfolio directory: `~/.dollhouse/portfolio/` on macOS and Linux, or `%USERPROFILE%\\.dollhouse\\portfolio\\` on Windows. The portfolio contains starter elements, user-created content, and installed community elements.

After connecting the server to an MCP client, you can describe an element in natural language, activate an existing item, or refine one that is already in the portfolio. Activation changes the AI's behavior or available capabilities for the session. Ensembles can activate several elements together and coordinate their activation and permission policies.

DollhouseMCP/mcp-server MCP server also includes a bidirectional converter for Dollhouse Skills and the later agent skills format. Converted skills can be managed as native Dollhouse elements or exported for platforms that use the other format.

## Setup and configuration

The interactive installer is started with:

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

It opens a browser-based setup wizard and can install the server for Claude Desktop, Claude Code, Cursor, VS Code, Codex, Gemini CLI, Windsurf, Cline, and LM Studio. The wizard detects existing installations and supports automatic updates or pinned versions.

Claude Code can also be configured directly:

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

For Claude Desktop, the project provides a desktop extension in `.mcpb` format that can be downloaded and opened. The server supports Windows, macOS, and Linux.

Permission-hook support varies by client. Claude Code has full native support; Cursor, VS Code, Codex, Gemini CLI, and Windsurf have partial support. Codex's partial support is Bash-only. Cline and LM Studio provide MCP setup in the documented release, while Claude Desktop has no native permission-hook path in that release.

## Tools and capabilities

The documented workflows expose capabilities to:

- Create and edit personas, skills, templates, agents, memories, and ensembles.
- Activate elements from the local portfolio.
- Search and browse available elements.
- Install elements from the community collection.
- Convert skills between Dollhouse and agent skills formats.
- Persist structured facts, preferences, and project state through memories.
- Apply permission policies through personas and ensembles.
- Run multi-step agent goals with state tracking and autonomy-related policies.

DollhouseMCP/mcp-server MCP server includes starter content and supports adding content created locally or obtained from the community collection.

## Limitations and notes

Core element management is documented as working across MCP-compatible platforms. Advanced Gatekeeper confirmation flows and agentic loop execution have been tested extensively on Claude Code and are expected to depend on standard MCP tool call and response behavior on other clients.

Permission hooks are not equally available across the supported clients, so installing the MCP server does not guarantee the same approval workflow everywhere. Skills introduced by DollhouseMCP predate the later agent skills format; conversion is provided for interoperability rather than treating the formats as identical.

The project is licensed under AGPL-3.0.

_Full upstream README: https://allmcps.com/mcp/dollhousemcp-mcp-server/readme_