Blender MCP Bridge
Drives Blender from an AI assistant by exposing Blender as MCP tools; the companion Blender add-on executes commands live in the scene. Enables creating primitives and materials, transforming/deleting objects, querying scene and object info, capturing viewport screenshots for visual feedback, and running arbitrary bpy Python (single commands or ordered batches) against the open .blend file.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Blender MCP Bridge
Drive Blender from any AI assistant over the Model Context Protocol. An MCP server (this repo) exposes Blender as tools; a Blender add-on executes them live in your scene.
MCP client (Claude Desktop / Cursor / Muse / ...)
--stdio--> server/blender_mcp_server.py
--TCP 127.0.0.1:9876--> Blender add-on --> bpy --> your sceneWhat you can do with it
"Create a red metallic sphere at (1, 0, 0)"→create_primitive+create_material+assign_material"What's in my scene?"→get_scene_info"Show me what it looks like"→get_viewport_screenshot(the AI sees the viewport)"Add a subdivision modifier and animate it rotating"→execute_blender_code(rawbpy)
Related MCP server: Blender Open MCP
Setup
1. Install the Blender add-on
Grab
dist/blender_mcp_bridge.zip(or zipaddon/blender_mcp_bridge.pyyourself).In Blender: Edit → Preferences → Add-ons → Install from Disk… → select the zip.
Enable Interface: Blender MCP Bridge.
Open the 3D Viewport sidebar (
N) → MCP Bridge tab → click Start Server.
The server listens on 127.0.0.1:9876. Blender must stay open while you use it.
2. Install the MCP server
cd blender-mcp
python -m venv .venv && source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt
python server/blender_mcp_server.py # sanity check: should start without errors3. Point your AI client at it
Claude Desktop (claude_desktop_config.json):
{
"mcpServers": {
"blender": {
"command": "/absolute/path/to/blender-mcp/.venv/bin/python",
"args": ["/absolute/path/to/blender-mcp/server/blender_mcp_server.py"]
}
}
}Cursor: Settings → MCP → Add Server → same command/args.
Muse CLI: Muse mcp add blender -- /absolute/path/to/.venv/bin/python /absolute/path/to/server/blender_mcp_server.py
Then ask your assistant to ping_blender — if it answers with your Blender version, you're live.
Tools
Tool | What it does |
| Health check; returns Blender version |
| Objects, materials, collections, render settings, frame range |
| Transforms, materials, bounds, mesh stats for one object |
| Viewport capture as an image (visual feedback loop) |
| Run arbitrary |
| Run multiple bridge commands in one round trip (ordered, per-item results) |
| cube / spheres / cylinder / cone / plane / torus / monkey / … |
| Move / rotate / scale by name |
| Remove an object by name |
| Principled BSDF material from hex color + metallic/roughness |
| Put a material on an object |
Transports: stdio (default) or HTTP
By default the server speaks MCP over stdio — what Claude Desktop, Cursor, and Muse CLI expect (see config above).
For setups where one server should be shared (multiple local clients without spawning a process each, or remote access via a tunnel), run it over Streamable HTTP instead:
python server/blender_mcp_server.py --http --port 8000
# serves MCP at http://127.0.0.1:8000/mcpPoint any Streamable-HTTP-capable client at that URL. Keep it on localhost unless you know what you're doing — the bridge executes arbitrary Python.
Protocol (v1)
TCP on 127.0.0.1:9876. Every message = 4-byte big-endian length + UTF-8 JSON.
Request:
{"id": str, "command": str, "params": dict}Response:
{"id": str, "ok": bool, "result": any, "error": str | null}
Commands are queued and executed on Blender's main thread via bpy.app.timers
(bpy is not thread-safe), then the response is written back on the same socket.
Multiple AI clients
Yes — connect as many as you like. Each AI client (Claude Desktop, Cursor, Muse CLI, …) spawns its own copy of the MCP server, and the Blender add-on accepts simultaneous connections from all of them. Commands are serialized through Blender's main thread, so concurrent clients interleave safely instead of corrupting the scene. The sidebar panel shows how many clients are currently connected.
One tip: two AIs editing the same scene will step on each other's toes creatively (both creating "Cube", etc.). It works, but give them different tasks — or different .blend files — for the smoothest experience.
Troubleshooting
"cannot reach Blender on 127.0.0.1:9876" → Blender isn't open, the add-on isn't enabled, or you didn't click Start Server.
Port already in use → another Blender (or a zombie process) holds 9876. Close it or restart Blender.
Screenshot fails → needs a visible 3D viewport; won't work in pure background (
-b) mode.
Security
The add-on executes arbitrary Python sent over localhost. That is its job — but only run it on machines you trust, and never forward port 9876 to a network.
Project layout
addon/blender_mcp_bridge.py # the Blender add-on (socket server + UI panel)
server/blender_mcp_server.py # the MCP server (FastMCP tools over stdio)
server/bridge_client.py # shared TCP client for the wire protocol
tests/ # protocol + threading tests (no Blender needed)
dist/blender_mcp_bridge.zip # installable add-on packageTool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
A comprehensive Model Context Protocol (MCP) server that enables AI assistants to control Unreal E…
Make games with AI: one MCP server that connects your agent to Unity, Unreal and Godot editors.
Generate game-ready 3D models, textures, and audio from natural language, over MCP.
A comprehensive Model Context Protocol (MCP) server that enables AI assistants to interact with yo…
Related MCP Servers
- AlicenseAqualityCmaintenanceEnables AI clients to control a running Blender instance through the Model Context Protocol, allowing scene inspection, object creation and editing, rendering, and execution of Blender Python. It works locally via a WebSocket bridge between the MCP server and a Blender add-on.15MIT
- FlicenseNot gradedqualityAmaintenanceEnables AI agents to inspect and control a live Blender session through MCP tools that create, modify, and delete objects, apply materials, render images, execute Python, and pull PolyHaven assets over a local TCP add-on. It also exposes a provider-agnostic LLM layer that can be pointed at Ollama, LM Studio, llama.cpp, OpenAI-compatible, or Azure endpoints and switched at runtime.-
- FlicenseNot gradedqualityBmaintenanceEnables AI agents to drive a live Blender session over MCP — inspecting scenes, creating and modifying objects, applying materials, rendering, and pulling PolyHaven assets — while routing natural-language prompts through a provider-agnostic LLM backend (Ollama, LM Studio, llama.cpp, OpenAI-compatible endpoints, or Azure) that can be switched at runtime.-
- FlicenseNot gradedqualityAmaintenanceEnables AI agents to drive a live Blender session over MCP by inspecting scenes, creating and modifying objects, assigning materials, rendering images, and fetching PolyHaven assets. It also exposes prompt-based scene building and review workflows backed by a provider-agnostic LLM layer (Ollama, LM Studio, llama.cpp, OpenAI-compatible endpoints, Azure) that can be switched at runtime.-