Blender / Bonsai (IFC) MCP servers and AI agent skills

How AI connects to Blender / Bonsai (IFC)

An MCP server or skill can only do what Blender / Bonsai (IFC) lets programs do. Blender / Bonsai (IFC) exposes 3 automation interfaces:

• Blender Python API (bpy) (Python): Almost everything in Blender is reachable from the bundled Python interpreter through the bpy module: scene data, operators, modifiers, materials and rendering. Every MCP server here works the same way, with an add-on inside Blender opening a local TCP socket and executing Python sent from the MCP process. Blender Lab's server also bundles API reference pages, so the agent can look up a bpy name before it writes code. Documentation
• IfcOpenShell and the Bonsai API (Python): Bonsai, formerly BlenderBIM, stores the building as a native IFC model handled by the IfcOpenShell Python library, and Blender objects are a view of that data. Edits made with ifcopenshell.api keep ownership, relationships and GlobalIds valid, which is why the IFC-aware servers expose an IfcOpenShell code tool separate from plain bpy. IfcOpenShell can also read and write .ifc files with no Blender running at all. Documentation
• Command-line and background mode (Command line): Blender runs without its interface using the -b flag and executes a script with --python, which suits batch jobs such as auditing many .blend files or rendering. Blender Lab's server has a matching set of tools that open a named file in a background Blender process instead of touching the session you have open. Documentation

What Blender Foundation offers

Blender Lab, which describes itself as the innovation space within the Blender project, lists its MCP Server as released. The server and add-on live on projects.blender.org under the GPL, need Blender 5.1 or newer, and name llama.cpp as the example client. Blender itself has no built-in LLM connection, and the Lab page warns that generated code runs without guards and recommends a virtual machine or a system without sensitive data.

Checked 4 Nov 2026. Sources: Blender Lab, MCP Server, Blender Lab projects page, lab/blender_mcp repository (projects.blender.org)

What you can do with Blender / Bonsai (IFC) and AI

• List every wall, door or space in an IFC model open in Bonsai, with property sets such as fire rating, filtered by class or storey.
• Produce a quantity takeoff of areas, volumes and lengths per IFC class and storey, or export a BC3 cost budget from the model.
• Read or set the georeferencing of an IFC file from an EPSG code before handing it to GIS or site coordination.
• Audit a .blend file for missing textures and linked libraries, or summarise its contents, in a background Blender without opening the interface.
• Build or adjust a 3D scene, apply materials, then capture viewport angles or renders so the agent can check its own geometry.
• Give the agent checked bpy, IfcOpenShell, Bonsai and Sverchok patterns through a skill pack so generated IFC entities keep valid relationships.

Taken from the listings' own documentation. Check each listing for exactly what it supports.

Installing

The listings here document setup for Antigravity, Claude Code, Claude Desktop, Codex, Cursor, Open WebUI (through the Docker REST/OpenAPI mode), OpenAI Codex, OpenCode, VS Code, Windsurf and llama.cpp. Install steps, copied from each project's own documentation, are on each listing's page. Most Blender / Bonsai (IFC) integrations run on your own computer next to Blender / Bonsai (IFC), so you need Blender / Bonsai (IFC) installed and licensed on the same machine unless the listing says otherwise.