blender-mcp
Allows LLMs to control Blender for creating and manipulating 3D scenes, objects, materials, shaders, modifiers, animations, and rendering.
Click on "Install 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., "@blender-mcpcreate a blue cube and set its material to glossy"
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.
blender-mcp
A Model Context Protocol server that lets LLMs such as Claude control Blender — creating scenes, manipulating objects, assigning materials, configuring shader nodes, applying modifiers, setting keyframes, and rendering.
┌──── HTTP (recommended) ────────────────┐
Claude Code ──────┤ ├──> Blender (GUI)
/ any client └──stdio──> launcher.py ──HTTP ───────────┘ └─ MCP Server on localhost:8400Blender must be running with the add-on enabled. The add-on exposes a Streamable HTTP MCP endpoint at http://localhost:8400/mcp — any MCP client that supports HTTP transport can connect directly. For clients that only support stdio transport, launcher.py acts as a thin stdio-to-HTTP proxy.
Prerequisites
Requirement | Notes |
Blender 4.0 or newer | Must run in GUI mode (not headless) |
Python 3.11+ (system) | Only needed for |
| Separate install — see below |
1. Install launcher dependencies (optional — stdio fallback only)
Skip this step if you are connecting via HTTP (recommended).
pip install httpx "mcp[cli]"2. Install mcp[cli] into Blender's bundled Python
Blender ships its own Python interpreter. The add-on runs inside it, so mcp[cli] must be installed there separately.
Find Blender's Python executable — its exact path depends on your OS and Blender version:
OS | Typical path |
Linux |
|
macOS |
|
Windows |
|
Then run:
# Linux / macOS
/path/to/blender/4.x/python/bin/python3.11 -m pip install "mcp[cli]"
# Windows (PowerShell)
& "C:\Program Files\Blender Foundation\Blender 4.x\4.x\python\bin\python.exe" -m pip install "mcp[cli]"Related MCP server: BlenderMCP
Getting the Add-on
Option A — Download a release (recommended)
Download blender_mcp_addon-vX.Y.Z.zip from the latest GitHub Release and install it directly in Blender (see Installing and Activating in Blender).
Option B — Clone the repository
git clone https://github.com/zorak1103/blender-mcp.git
cd blender-mcpThe add-on directory is blender_addon/. No build or compilation step is needed — Blender add-ons are plain Python packages.
Option C — Build the zip locally
Blender can install add-ons from a zip file containing the package directory.
Requires Hatch (pip install hatch):
hatch run packageOr without Hatch:
zip -r blender_mcp_addon.zip blender_addon/Installing and Activating in Blender
From a zip file
Open Blender and go to Edit → Preferences → Add-ons.
Click Install… and select
blender_mcp_addon.zip.Search for "Blender MCP Server" and enable the checkbox.
From the cloned directory
Open Blender and go to Edit → Preferences → Add-ons.
Click Install…, navigate into the repository, and select the
blender_addonfolder itself (or its__init__.py).Search for "Blender MCP Server" and enable the checkbox.
Verify activation
After enabling, the Blender Info bar (top of the screen) should show:
Blender MCP Server registered, port=8400Change the port (optional)
In Edit → Preferences → Add-ons → Blender MCP Server, set the Port field (default: 8400) before enabling, or disable/re-enable after changing it.
Connecting an MCP Client
Important: Blender must already be running with the add-on enabled before connecting.
Option A — Direct HTTP (recommended)
The add-on exposes a Streamable HTTP endpoint at http://localhost:8400/mcp. Any MCP client
that supports HTTP transport can connect directly — no proxy, no extra Python dependencies.
The MCP endpoint requires a Bearer token that the add-on generates on first start and writes to
~/.config/blender-mcp/token. Export it as an environment variable before launching your client:
# bash / Git Bash / WSL (add to ~/.bashrc for persistence)
export BLENDER_MCP_TOKEN="$(cat ~/.config/blender-mcp/token)"# PowerShell — set permanently in the user environment
[Environment]::SetEnvironmentVariable(
"BLENDER_MCP_TOKEN",
(Get-Content "$HOME\.config\blender-mcp\token"),
"User")Claude Code — the .mcp.json at the repository root is already configured with the
Authorization: Bearer ${BLENDER_MCP_TOKEN} header. Set the env var above, then start
(or restart) Claude Code — no further setup is needed inside this repository.
To register globally (outside this repo):
claude mcp add --transport http --scope user \
--header "Authorization: Bearer \${BLENDER_MCP_TOKEN}" \
blender-mcp http://localhost:8400/mcpOther clients (Cline, OpenCode, etc.): point the HTTP/SSE transport URL to
http://localhost:8400/mcp and add the Authorization: Bearer <token> header.
Note for Claude Code users: MCP servers from
.mcp.jsonrequire either"enableAllProjectMcpServers": truein.claude/settings.local.json, or explicit approval via the dialog that appears on first launch. If the tools are not available and no dialog appeared, add"enableAllProjectMcpServers": trueto.claude/settings.local.jsonand restart Claude Code.
Option B — stdio proxy (fallback)
If your MCP client only supports stdio transport, use launcher.py as a thin proxy.
This requires system Python with httpx and mcp[cli] installed (see Prerequisites).
Add the following to ~/.claude/settings.json (global) or .claude/settings.json (project-level):
{
"mcpServers": {
"blender-mcp": {
"command": "python",
"args": ["/absolute/path/to/blender-mcp/launcher.py"]
}
}
}launcher.py will wait up to 60 seconds for Blender to become reachable, then time out.
Verify the connection
# HTTP check (works for both Option A and B):
export BLENDER_MCP_TOKEN="$(cat ~/.config/blender-mcp/token)"
curl -s -X POST http://localhost:8400/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "Authorization: Bearer $BLENDER_MCP_TOKEN" \
-d '{"jsonrpc":"2.0","method":"initialize","id":1,"params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"0"}}}'# stdio proxy round-trip (Option B only):
echo '{"jsonrpc":"2.0","method":"tools/list","id":1}' | python launcher.pyTroubleshooting
MCP tools not available in Claude Code / approval dialog never appeared
Claude Code loads servers from .mcp.json only when enableAllProjectMcpServers is true or the user approves them via a one-time dialog. If neither happened, the server silently stays inactive.
Fix: add or update .claude/settings.local.json in the repository root:
{
"enableAllProjectMcpServers": true
}Then restart Claude Code. The blender-mcp tools should appear immediately.
curl http://localhost:8400/mcp returns 401 Unauthorized
The MCP endpoint requires a Bearer token. Make sure BLENDER_MCP_TOKEN is set and matches the content of ~/.config/blender-mcp/token (written by the add-on on first start):
export BLENDER_MCP_TOKEN="$(cat ~/.config/blender-mcp/token)"Then restart Claude Code (or whichever client you are using) so it picks up the new environment variable.
curl http://localhost:8400/mcp returns "Not Acceptable"
This is expected and means the server is running. The endpoint requires specific headers. Use the full POST check from Verify the connection instead.
The add-on appears enabled but tools/list returns an error or times out
Confirm the Info bar shows
Blender MCP Server registered, port=8400after enabling the add-on.If you changed the port, update the URL in
.mcp.jsonaccordingly.Try disabling and re-enabling the add-on in Edit → Preferences → Add-ons.
mcp[cli] not found when enabling the add-on in Blender
The package must be installed into Blender's own Python, not the system Python. Re-run the install command using the Blender Python executable (see Prerequisites).
Available Tools
Category | Tools |
Scene |
|
Objects |
|
Materials |
|
Render |
|
Shader nodes |
|
Modifiers |
|
Animation |
|
Lighting |
|
Camera |
|
World |
|
Scripting (opt-in) |
|
Further Reading
Topic | File |
Security & | |
Development setup, testing, releasing | |
LLM agent self-configuration guide |
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/zorak1103/blender-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server