Blender Foundation

Blender Foundation deployed an experimental MCP (Model Context Protocol) server as part of the Blender Lab initiative, offering users a way to control Blender using natural language commands sent through an external AI assistant such as Claude. The add-on requires Blender 5.1 or newer and is available as an opt-in experimental feature.

The Blender Lab MCP server is a lightweight add-on that implements the Model Context Protocol, an open standard that lets large language models (LLMs) connect to external tools. Users install a Blender add-on that opens a local TCP connection; an external LLM client (such as Claude Code or Cursor) then sends natural language instructions, which are translated into Blender Python API commands and executed inside the running Blender session. The server can inspect scenes, execute Python code, trigger renders, and navigate the interface. The project repository was first committed in February 2026 and a public page was published at blender.org/lab/mcp-server/.

Blender integrated AI-based denoising into its Cycles render engine, offering users machine learning tools that clean up noisy rendered images, significantly reducing the number of render samples needed. Users can opt to apply Intel's Open Image Denoise (OIDN) or NVIDIA's OptiX AI denoiser within the software.

The Denoise node in Blender's compositor uses Intel's Open Image Denoise (OIDN), which applies machine learning to transform noisy rendered images into clean outputs. OIDN was first added in Blender 2.81 (2019), and NVIDIA OptiX AI-Accelerated viewport denoising was added in Blender 2.82 (2020). Both denoisers take noisy render passes (color, albedo, normal) as input and produce cleaned image output. As of Blender 4.1, OIDN can also run on the GPU across NVIDIA, AMD, and Apple Silicon hardware.