Persona Motion Studio
Enables AI assistants to show, animate, and control a VRM character on the desktop, including posing and motion installation via MCP tools.
README
Persona Motion Studio
A local desktop home for a VRM character that an MCP-enabled AI assistant can show, animate, and react through — with a visual motion editor built in.

What it does
- imports your own humanoid
.vrmcharacter into a private on-device library; - keeps the character on top of your desktop while you work;
- lets you pose body, hands, and individual fingers with degree controls;
- creates, edits, duplicates, scrubs, previews, and saves animation keyframes;
- exports portable
.vrmafiles and editable.persona-motion.jsonprojects; - installs named motions into Persona so an AI assistant can discover and play them through local MCP tools; and
- reacts to supported desktop voice output without recording or saving audio.
No character model is included in the repository or installer.
Install on Windows
- Open the repository's Releases page.
- Download
Persona-Motion-Studio-...-windows-x64-setup.exe. - Run the installer and open Persona Motion Studio.
- Choose a
.vrmcharacter when the welcome guide appears.
The installer contains the native audio helper. Your friends do not need Node.js, Git, Visual Studio, or this source repository.
Windows may show a SmartScreen warning for early unsigned builds. Check that
the download came from this repository's release page and compare its SHA-256
value with SHA256SUMS.txt before continuing.
Linux AppImage/DEB and macOS DMG/ZIP packages are built by the release workflow. The first beta is primarily tested on Windows.
Five-minute walkthrough
1. Import a character
Press Import your VRM and choose a humanoid .vrm file. Persona copies it
into the app's private data folder. It is not added to projects, source control,
MCP responses, or network requests.
Only use a model whose creator permits your intended use. Some VRoid Hub models allow personal avatar use but forbid redistribution or modification.
2. Make a pose
Choose a bone in the left panel. Move its X, Y, and Z controls in the right panel. The selected joint is marked on the character, and Show skeleton overlay makes the whole rig easier to understand.
Hands and individual thumb, index, middle, ring, and little-finger joints are available when the imported VRM exposes those standard humanoid bones.
3. Turn poses into movement
Select Frame 1, adjust the pose, and press Save changes to Frame 1. Duplicate the frame, give it a later time, change the pose, and save again. Press Play to preview the interpolation between frames.
The editor autosaves locally. Save project creates an editable project file you can share without sharing the character. Export VRMA creates a portable animation.
4. Install it for your assistant
Give the motion a memorable name, choose a reaction slot, and press Install for GPT. Persona stores the motion in the same private local library and previews it immediately.
5. Connect Codex or another MCP client
Leave Persona running and add its loopback-only MCP endpoint:
codex mcp add persona --url http://127.0.0.1:47831/mcp
Start a new Codex task after adding the connection. Ask it to list saved Persona motions, play one, open Motion Studio, or show/hide the character.
Other desktop AI clients can use the same Streamable HTTP endpoint if they
support local MCP servers. A hosted web-only assistant generally cannot reach
127.0.0.1 on your computer.
Privacy and security
- The VRM and custom motion library stay under the operating system's app-data folder.
- Release builds actively remove
model.vrmand local custom-motion folders even if a developer accidentally leaves them in the source tree. - MCP exposes six narrow Persona tools, not filesystem or shell access.
- The local server binds only to
127.0.0.1and rejects non-loopback host headers. - Voice activity is calculated from supported application output in memory. Persona does not capture a microphone, transcribe, save, or upload audio.
See Integrations for the exact tool contract.
Build from source
Requirements:
- Node.js 24 or newer;
- npm;
- hardware-accelerated desktop graphics; and
- Visual Studio Build Tools with Desktop development with C++ for Windows packaging.
npm install
npm run check
npm run demo
Create an installer for the current operating system:
npm run dist:windows
# or: npm run dist:linux / npm run dist:mac
Build outputs go to release/. Character media and user-created VRMA files are
ignored by Git and excluded from the package.
Project lineage and licence
This project is based on xikhar/persona, whose excellent local desktop character and voice-reactive foundation is MIT-licensed. See Third-party notices.
Application source and code-generated starter motions are covered by MIT. Imported characters retain their creators' licences and are not covered by this repository's licence.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。