M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
B

Blender MCP Server

Community
Let any LLM control Blender 3D scenes in real time via structured JSON commands
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #374 of 438
Stars
★ 98 Popular
Runtime
Node.js · Python
Credentials
No credential needed
License
MIT
Last commit
⚠
37FMRS · D

An MIT-licensed community MCP server that bridges LLMs and Blender via a JSON-over-TCP protocol, offering real-time scene control and LLM-agnostic compatibility. The README does not enumerate specific tools or provide a client config example, so some details require checking its online documentation. Well suited for natural-language Blender workflows; not suitable for those needing official support or non-Blender pipelines.

Strongest · Setup experience 9/20 Weakest · Reliability 5/20

Reliability
5/20
Security and permissions
8/20
Maintenance
9/20
Documentation
6/20
Setup experience
9/20
Why each score
Reliability 5/20
No server manifest was available (marked 'not cached'), and the README lists no concrete MCP tools, parameters, or response formats, so declared tool behavior cannot be checked against real behavior. No test files or CI workflow evidence are shown in the provided material; 'node index.js' is the only startup instruction, with nothing about init handshake, error handling, or dependency control. Per the static calibration rule, reliability cannot exceed 12 without verifiable execution evidence (real CI plus committed tests covering key paths); given the near-total absence of such evidence, a low score is warranted.
Security and permissions 8/20
The README describes the server communicating with Blender over a TCP socket using JSON, but says nothing about whether the port is bound to localhost only, whether authentication exists, or whether dangerous operations (e.g., deleting scene data, writing files, executing arbitrary scripts) require confirmation or are scoped. There is no evidence of credential theft, malicious code, or a known unmitigated vulnerability, so this does not trigger a red-line (not scored 0-4), but the complete lack of disclosure on permission scope, data boundaries, and confirmation flows for potentially destructive actions is a significant gap, warranting a substantial deduction.
Maintenance 9/20
The repo is MIT-licensed, not archived, and has 96 stars with 0 open issues, indicating some visibility and apparent upkeep. However, the provided material gives no commit history, release/versioning record, issue-response cadence, or dependency-update evidence, nor any clear maintainer identity or security-response channel, so it cannot be rated as 'sustained maintenance with clear ownership and update path' — it lands at 'active but with governance/versioning gaps.'
Documentation 6/20
The README gives basic install steps (clone, pnpm install, node index.js) and a link to an external documentation site, but the in-repo documentation does not cover tool parameters, authentication, usage limits, cost, or troubleshooting; the described package manager (pnpm) is also unverifiable against the actual repository structure from this evidence alone. Because key usage details are pushed off to an external site rather than reviewable in-repo content, documentation quality is only 'usable but with hidden assumptions and troubleshooting gaps.'
Setup experience 9/20
The install steps are simple and low-friction (clone, install deps, run script), which is a positive. However, there are no example configurations for connecting from a mainstream MCP client (e.g., Claude Desktop), no stated port/environment variables, no way to verify a successful connection, and no CI or test evidence that this flow actually works end to end. Per the static calibration rule, setup cannot exceed 15 without real CI and tests covering key paths, and the missing client-integration examples and verification steps warrant a further deduction.

Static review · not runListed 2026-09-07

Read the FMRS scoring method →

Fit and risk

What it can accessRuns commands or codeUses the network

Best for

  • 3D artists who want real-time LLM-driven scene control in Blender
  • Developers and hobbyists driving Blender via natural language
  • Users needing an LLM-agnostic integration (GPT, Claude, local models)

Not for

  • Users who do not work with Blender
  • Those seeking an officially supported Blender tool (this is community-built)
  • Users who need a GUI-driven workflow rather than command-based control

Required permissions

  • Requires network/socket access to send JSON commands over TCP
  • Requires read/write control over Blender scenes and objects

Risks and side effects

  • LLM-generated commands may unintentionally modify or damage the current Blender scene
  • TCP socket communication could be exposed on the local network if not secured
  • The README does not document authentication, so security hardening must be assessed independently

Setup

Before you start

Runtime：Node.js · Python

  1. Clone the repository: git clone https://github.com/pranav-deshmukh/blender-mcp.git and cd into it
  2. Install dependencies: pnpm install
  3. Start the MCP server: node index.js. Docs and demo at https://blender-mcp-studio.vercel.app/.

Check that it works

Start the server with node index.js and add it to your MCP client config; confirm a Blender-related tool appears in the client's tool list, or ask "list all objects in the current scene" to prove the real-time TCP connection to Blender works.

Troubleshooting

  1. Verify all dependencies were installed with pnpm install
  2. Check that node index.js starts the server without errors
  3. Confirm Blender is running and its connection settings match the TCP socket configuration
  4. Consult the usage docs at https://blender-mcp-studio.vercel.app/

Things to try

Once connected, you can ask your AI assistant things like:

  • Connect to Blender and list all objects in the current scene
  • Create a cube in the scene and move it to the origin
  • Add a material to the selected object
  • Rotate the viewport and capture the current 3D scene

Use cases

Control Blender 3D scenes with natural language
Full scene and object manipulation from code
Integrating Blender into MCP-enabled LLM workflows

Overview

Blender MCP is a plug-and-play Model Context Protocol server that lets any Large Language Model (GPT, Claude, or even local models) interact with Blender in real time using structured JSON commands over a TCP socket, enabling natural language control of 3D scenes. Maintained by pranav-deshmukh under the MIT License, it is a community project and not an official Blender product.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 3d2219f3f89e Data synced 2026-10-19 Read the FMRS scoring method