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

Houdini MCP Server

Community
Connect SideFX Houdini to Claude via the Model Context Protocol.
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #419 of 438
Stars
★ 305 Popular
Transport
stdio (local process)
Runtime
Python
Credentials
Optional API key
License
MIT
Last commit
28FMRS · D

This project is based on blender-mcp, providing local control of Houdini from Claude. It is feature-rich but requires manual setup and is not an official product. The documentation is clear, but users should be cautious about permissions and risks.

Strongest · Setup experience 8/20 Weakest · Security and permissions 2/20

Reliability
4/20
Security and permissions
2/20
Maintenance
6/20
Documentation
8/20
Setup experience
8/20
Why each score
Reliability 4/20
Only the README and repository metadata were supplied; no server source, tests, or CI configuration are available. The MCP initialization handshake, tool list, and real tool behavior cannot be verified. The README describes a plausible happy path, but without reproducible execution evidence the score must be conservative.
Security and permissions 2/20
The README describes a Houdini plugin listening on localhost:9876 and handling commands such as creating/modifying nodes and executing code, with no mention of authentication, authorization, or user confirmation. This is effectively an unauthenticated local command-execution service that any local process could trigger, an unmitigated remote-command default risk. This triggers the red line, so security is scored at 2.
Maintenance 6/20
The repository is not archived and has an MIT license, but there is no evidence of commit history, releases, dependency updates, a security response channel, or issue responsiveness. Stars and open-issue counts are discovery signals only and do not count toward the score. Maintenance activity is therefore unverified and scored conservatively.
Documentation 8/20
The README includes installation steps, uv usage, Claude Desktop/Cursor configuration, and optional OPUS integration, but it lacks tool parameter references, functional limitations, cost details, and testing guidance. The table of contents promises Testing & Usage and Troubleshooting sections that are not present in the supplied content. Documentation is usable but has hidden assumptions and notable gaps.
Setup experience 8/20
The setup instructions are fairly detailed, covering folder layout, a Shelf Tool, packages integration, and Claude Desktop configuration examples. However, the process is manual, Windows-centric, and relies on copying files and configuring paths by hand. No automated verification, CI workflows, or committed tests were provided. Under the static calibration, setup cannot exceed 15 without verifiable execution evidence, and none is present.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessRuns commands or codeUses the network

Best for

  • Houdini artists and technical directors who want a seamless workflow between Claude and Houdini.
  • Users interested in AI-assisted procedural 3D modeling and scene generation.
  • Developers familiar with MCP who want to extend Claude to work with 3D tools.

Not for

  • Users without a Houdini license (requires SideFX Houdini).
  • Users seeking official support (this server is not maintained by SideFX or Anthropic).
  • Users wanting a remote server without local Houdini access (the server runs on localhost).

Required permissions

  • Local network access: listens for TCP connections on localhost:9876.
  • File system access: reads and writes files in the Houdini configuration directory.
  • Execution permission: executes Python code within the Houdini context.

Risks and side effects

  • Executing untrusted code in Houdini can lead to data loss or system damage; only use in trusted environments.
  • Running a server on localhost may expose it to other local processes; ensure no malicious software.
  • OPUS integration requires a RapidAPI key, which if mishandled could leak sensitive information.

Setup

Before you start

Runtime：Python

  1. Install Houdini and uv.
  2. Create a plugin folder in your Houdini scripts directory (e.g., C:/Users/YourUserName/Documents/houdini19.5/scripts/python/houdinimcp/) and place __init__.py, server.py, and pyproject.toml inside.
  3. (Optional) Create a shelf tool or packages integration to auto-load the plugin.
  4. In the plugin directory, add MCP using uv: cd ... && uv add "mcp[cli]".
  5. Configure Claude Desktop or Cursor by adding an MCP server entry pointing to houdini_mcp_server.py.
  6. If uv run succeeds but Claude fails to load MCP, ensure Claude is using the same Python version and replace "python" in the config with the path from python -c "import sys; print(sys.executable)".
  7. (Optional) Set up OPUS integration by copying urls.env.example to urls.env and adding your RapidAPI key.
claude_desktop_config.json 
{
  "mcpServers": {
    "houdini": {
      "command": "uv",
      "args": [
        "run",
        "python",
        "C:/Users/<YourUserName>/Documents/houdini19.5/scripts/python/houdinimcp/houdini_mcp_server.py"
      ]
    }
  }
}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "houdini": {
      "command": "uv",
      "args": [
        "run",
        "python",
        "C:/Users/<YourUserName>/Documents/houdini19.5/scripts/python/houdinimcp/houdini_mcp_server.py"
      ]
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add houdini -- uv run python 'C:/Users/<YourUserName>/Documents/houdini19.5/scripts/python/houdinimcp/houdini_mcp_server.py'

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

Start Houdini and toggle the MCP server with the shelf tool (localhost:9876), then confirm the houdini server's tools appear in Claude Desktop's tool list; a simple node-creation request via Claude proves the connection works.

Troubleshooting

  1. If the plugin doesn't start, check that the Houdini Python path includes the plugin directory.
  2. If Claude fails to load MCP, verify uv is in PATH and that Claude's Python version matches the uv environment.
  3. If connection fails, ensure the Houdini MCP server is running (via shelf tool or package) and port 9876 is not occupied.
  4. If OPUS tools are disabled, check if urls.env exists and contains a valid RapidAPI key.

Things to try

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

  • Create a geometry node network in my Houdini scene
  • Modify the parameters of the currently selected Houdini node
  • Execute a snippet of Python code inside Houdini
  • Load a furniture asset from OPUS into the scene

Use cases

Control Houdini from Claude using natural language to create and modify nodes.
Execute Python code inside Houdini to automate tasks.
Leverage OPUS integration to generate furniture and environmental procedural assets.

Supported clients

Claude Desktop
Cursor

Listed from the project's documentation, not tested by this site.

Overview

HoudiniMCP is an MCP server that allows you to control SideFX Houdini from Claude. It consists of two parts: a Houdini plugin (Python package) that listens on a local port (default localhost:9876) and handles commands (creating and modifying nodes, executing code, etc.), and an MCP bridge script that communicates with Claude via stdin/stdout and with Houdini via TCP. The server requires uv and Houdini, and offers optional OPUS integration for furniture and environmental procedural assets.

Similar servers

Context7 80 · B

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

★ 62.9k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision de4fd93acc20 Data synced 2026-10-13 Read the FMRS scoring method