Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@OpenBody2039 MAYA MCP Check MAYA's status, then have her stand up and wave at me."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Give AI Agents a Body. Her name is MAYA.
Part of MAYA 3.0 — the end-to-end IoT agent stack from LEGEND.
Landing: live · Blog: www.ura2039.xyz · repo
MAYA 3.0 is our end-to-end IoT agent: from cloud mind → MCP body knowledge → live avatar motion.
This repo ships the OpenBody2039 MAYA MCP body layer. Product vision, robotics notes, and the full agent story live on www.ura2039.xyz — 热热的博客 / LEGEND.
Universal Model Context Protocol server for avatar body knowledge and intent-level control.
MAYA 3.0 IoT agent — embodiment for agents that already think
IDE-agnostic stdio MCP (Claude Desktop, Cursor, Cline, Continue, custom agents, …)
Self-contained knowledge in body-graph/ (works offline)
Optional live control via local WebSocket bridge + any OpenBody-compatible viewer
Learn more on the LEGEND blog: www.ura2039.xyz
Product | MAYA 3.0 end-to-end IoT agent |
Blog | |
Node | ≥ 20 |
OS | Linux (Ubuntu 22.04 tested) / macOS should work |
Deps |
|
License | MIT |
Agents need more than chat. MAYA 3.0 connects reasoning to a constrained, performable body — joints, expressions, intent primitives — so an IoT agent can wave, nod, stand, and stay alive in the scene.
Follow the roadmap and robotics write-ups on URA2039 (ROS2 · Linux · Python · agents).
Related MCP server: radoss_avatar
npm install
npm run export:graph # rebuild body-graph from avatar-source/
npm run test:mcp-offline # no viewer required
npm run mcp # stdio MCP (stdout = protocol only)Optional live stack:
npm run bridge # ws://127.0.0.1:8787
# open any OpenBody-compatible viewer with mcp mode enabled (often ?mcp=1){
"mcpServers": {
"openbody2039-maya": {
"command": "node",
"args": ["/ABS/PATH/to/OpenBody2039_MAYA/mcp/server.js"],
"env": {
"OPENBODY_WS_URL": "ws://127.0.0.1:8787"
}
}
}
}See docs/MCP.md and mcp/clients.example.json.
Tool / resource | Offline | Role |
| yes | Manifest |
| yes | Joints / axes / morphs / poses |
| yes | Full graph resources |
| needs bridge | Live control |
Prefer intent / primitives over joint-level tweaks. Bone names only from body_lookup.
Surface | Link |
MAYA 3.0 IoT agent story & blog | |
This MCP landing (GitHub Pages) | |
Source |
mcp/ stdio MCP server
brain/ intent → primitives
transport/ optional WS bridge process
body-graph/ exported knowledge (do not hand-edit)
avatar-source/ SSOT for export:graph
docs/ MCP / protocol / layers + landing
skill/ optional Cursor Agent Skill (NOT required by MCP)
generator/ export + smoke testsIn scope: protocol, body graph, MCP tools/resources, bridge — the body layer of MAYA 3.0.
Out of scope (this repo): full 3D avatar product binaries, GLB assets, Mixamo/AMASS motion corpora, proprietary viewer UI.
Product narrative and IoT agent architecture: www.ura2039.xyz.
Compatible viewers implement docs/PROTOCOL.md themselves.
Variable | Default |
|
|
|
|
|
|
|
|
MIT — see LICENSE.
MAYA 3.0 · Give AI Agents a Body · Blog www.ura2039.xyz · We are LEGEND.
This server cannot be deployed
Governed personal world model and memory for your AI agent. Pair once, connect over MCP.
Your personal world model for any agent.
Persistent memory, hybrid search and a goal graph for AI agents, over stdio or remote HTTP.
3D avatars, embeds, glTF tools, agent memory, and on-chain agent identity from three.ws.
