Experimental — Work in Progress

Your virtual world,
powered by Agentic AI

A Docker stack for OpenSimulator that integrates MCP servers, a bot client, Blender, Opencode and pretty much any AI provider or model you want to use.

You can build, manage, and explore your grid, all aided by a voice capable bot. Your assistant can spawn other more specialised bots. You can even create new 3D assets and import them into the world, and iterate in-world from natural-language workflows.

Get Started View on GitHub
OpenSimulator Chat 
Follow me

Bow before me, peasant!

Create a cube prim next to me and make it red.

Fly to me, then land and jump 3 times

Who is online right now?

Restart the region

Create new user 'Joe Smith' with password 'password123' 
and give them the 'Estate Manager' role.
Import OAR file from https://example.com/myregion.oar
and load it into this region

Kick user 'The Troll' from the grid and ban them.

Make the sky some crazy colours!

Go sit on the nearest chair and wave to everyone in the region.

How many prims are in this region and how many are owned by me?

Write an LSL script to make a prim spin & change colors randomly.

Create a group 'Bots Anonymous' and invite 'Kryten 2X4B-523P'.
Import IAR file from https://example.com/myinventory.iar 
and load it into my inventory.

Subdivide this parcel into 4 smaller parcels and set the 
access to 'Group Only' for each.

Follow me to the nearest teleport point and then teleport 
to 'Main Plaza'.

How many pairs of shoes are in my inventory? List their names.

Are there any scripts running in this region that are owned 
by other users? If so, list them.

Features

Everything you need to run an AI-enabled OpenSimulator grid in one Docker stack.

Agentic Behaviour

The AI doesn't just answer questions — it plans, decides, and acts. It can chain together console commands, in-world actions, and external tools to complete multi-step goals on its own.

Natural Language Control

Manage regions, users, and configuration through conversation. Anything the OpenSimulator console can do, your AI can do.

Web Interface

Use the web interface to manage the stack, including users, simulators, and bots. It also includes a simple user approval system and an easy setup wizard.

In-World AI Bot

A bot avatar logs in and accepts IM instructions. It can walk, run, fly, teleport, build prims, and manage inventory.

Voice-Ready Bot

Give your AI bot a voice and let it speak in-world with low-latency Piper synthesis streamed via Janus WebRTC.

Support for speech-to-text is planned.

Build & Create

Create, position, scale, rotate, and texture prims. Configure environment settings and upload media — all through the bot.

MCP Bridges

Three MCP servers power the stack: REST console admin, in-world bot control, and Blender authoring tools. The spawner adds bot lifecycle control on top.

Docker Simplicity

One compose file brings up OpenSimulator, MariaDB, MCP sidecars, Opencode, and opensim-spawner. Bots run in their own containers with resource limits.

Flexible AI Providers

Switch models and providers at runtime with star commands. Use GitHub Copilot, local models, or any provider Opencode supports.

The Stack

Open-source components wired together to create an AI-native virtual world.

OS

OpenSimulator

The open-source virtual world server that powers your regions, avatars, and assets.
OC

Opencode

AI agent runtime that connects to MCP servers and routes natural-language requests to tools.
LM

LibreMetaverse

The .NET library that lets the bot client log in, move, build, and interact with the grid.
C2

opensim-console2mcp

Bridges the OpenSimulator REST console to MCP so AI can run admin commands.
M2

opensim-metaverse2mcp

Bot-side MCP server that exposes in-world tools for movement, building, and inventory.
SP

opensim-spawner

Creates and manages multi-bot AI runtime instances on demand, including isolated opencode and metaverse2mcp containers per bot.
BL

opensim-blender

Blender MCP bridge container for AI-assisted 3D authoring workflows, powered by blender-mcp.
PI

opensim-piper

Voice synthesis sidecar that powers bot speech with Piper HTTP endpoints for TTS and voice discovery.
JA

opensim-janus

Janus WebRTC server container used as a central voice-transport component for in-world voice support.
B

Bot Client

The in-world avatar that receives IMs, executes tool calls, and acts as your grid assistant.

Architecture Diagram

Core runtime with on-demand multi-bot containers, MCP bridges, and the voice transport path in one view.

flowchart TB
  User[Viewer user] -->|IM commands| Bot[Bot avatars in OpenSimulator]
  Blender[Blender desktop app] -->|Author and export GLB/GLTF| OpenSim[opensim container]

  subgraph MCP["MCP Bridge Side"]
    Console[opensim-console2mcp]
    BlenderMCP[opensim-blender via blender-mcp]
  end

  Spawner[opensim-spawner]
  Piper[opensim-piper]
  Janus[opensim-janus]

  subgraph BotRuntime["Per-bot runtime (0..N)"]
    MetaverseN[opensim-metaverse2mcp]
    OpencodeN[opensim-opencode]
    MetaverseN <-->|MCP tools| OpencodeN
  end

  OpenSim -->|REST console| Console
  Spawner -->|Create/delete bot runtimes| MetaverseN
  Spawner -->|Create/delete bot runtimes| OpencodeN
  MetaverseN -->|Bot login and world actions| OpenSim
  BlenderMCP -->|Drive Blender tools| Blender
  MetaverseN -->|TTS over HTTP| Piper
  MetaverseN -->|WebRTC voice session control| Janus
  Piper -->|Synthesized audio stream| Janus
  Janus <-->|WebRTC voice transport| Bot
  Bot -->|Voice replies| User

  OpencodeN <-->|MCP tools| Console
  OpencodeN <-->|MCP tools| BlenderMCP
  Provider[AI provider/model] <-->|Model calls| OpencodeN

Ready to deploy your own grid?

Get the stack running in minutes with Docker Compose, then log in and start talking to your AI bot.

Get It Play It