Doug-218/blender-ai-copilot-forge

Overview

blender-ai-copilot-forge is a Blender extension that acts as a scene-building copilot. You describe a layout in natural language—for example a pedestal, a camera, three lights and a matte blue material—and the add-on reads the existing scene to plan the work, then forges primitives, duplicate rows, cameras, lighting, materials and parenting as ordered operations that you approve one at a time. It targets 3D artists who block out scenes by hand and want to skip repetitive menu and transform work.

The copilot runs inside Blender with a sidebar panel ("MCP Pro") and a Studio window showing chat, scene objects, viewport snapshots and a task board. Larger briefs can be delegated to a manager–worker–reviewer agent chain, and external MCP clients can reach the same toolset through a local authenticated endpoint at http://127.0.0.1:9876/mcp How it works.

Quick start

1. Download mcp-for-blender-pro.zip from Releases.
2. In Blender, open Edit → Preferences → Get Extensions → menu → Install from Disk and select the ZIP.
3. Enable mcp-for-blender-pro, then open 3D Viewport → N → MCP Pro.
4. Choose a model provider, enter the connection details and select a model.
5. Send a message in Blender, or click Open Studio Window, and approve each proposed build step.

The README gives a sample brief: "Create a simple product-display scene with a pedestal, a camera and three lights—give the pedestal a matte blue material" Usage.

Providers and settings

Workflow mode is either a single assistant or a Team chain. Supported providers include OpenAI, Anthropic Claude, Google Gemini, OpenRouter, OpenCode Zen, Ollama and a custom API; custom endpoints can use OpenAI Chat Completions, Anthropic Messages or the Ollama protocol. The MCP endpoint is bound to localhost by default and uses a generated bearer token. Approvals default to per operation, and the interface language can be English, German, French, Spanish or Chinese Configuration.

Constraints

The badges state Windows x64 and Blender 4.2+ Blender requirement, so other operating systems are not documented as supported. A model provider with tool calling is required, and usage depends on that provider's API access and costs. Because execute_code is one of the supervised tools, the approval step is the main safeguard before generated code runs in Blender. Exports via export_scene write to the path and format named in the brief.

HotRepo's take

The project entered HotRepo's new-project list on 2026-11-14 with 66 stars gained over roughly 11.5 hours of observation, reaching 167 stars and 4 forks; it has no recorded all-language daily rank. That growth is early and comes with almost no code visible in the repository, so the extension itself is effectively only available as the prebuilt ZIP in Releases—worth checking release history and issue activity before relying on it in a production pipeline.