blender-ai-mcp-addon: Visual AI Chat Inside Blender Anujasid's blender-ai-mcp-addon, a free MIT-licensed Blender addon that embeds an AI chat panel in the 3D Viewport's N-sidebar, reached nearly 2,000 GitHub stars in four days after its November 12 release. The addon inverts the standard Blender MCP setup by keeping the chat inside Blender rather than driving the app from Claude Desktop, Cursor, or a terminal, and ships 21 scene tools plus a Team chain for multi-agent manager, worker, and reviewer workflows. It supports OpenAI/ChatGPT, Anthropic Claude, Google Gemini, Ollama, OpenCode Zen, and custom OpenAI-, Anthropic-, or Ollama-format endpoints, with Ollama enabling offline local inference at no per-token cost. A Blender addon that puts an AI chat panel inside the 3D viewport — not in a separate app — went from zero to nearly 2,000 GitHub stars in four days. The blender-ai-mcp-addon https://github.com/anujasid/blender-ai-mcp-addon by anujasid dropped November 12 and is now trending. It is free, MIT-licensed, and it flips the standard Blender MCP setup: instead of driving Blender from Claude Desktop, Cursor, or a terminal, the chat interface lives inside Blender itself. Why This Is Different From Every Other Blender MCP Setup The existing approach — mcp-for-blender https://github.com/ahujasid/mcp-for-blender 34,600 stars being the most popular — works by opening a socket server inside Blender that an external MCP client then drives. You type in Claude Desktop, Blender moves. That works, but it splits your attention between two windows. blender-ai-mcp-addon inverts this. The addon adds a panel to the 3D Viewport's N-sidebar. Chat happens there. For more complex work, you open the Studio Window: a companion panel showing your conversation history, a live scene objects list, viewport screenshots, an approvals queue, and a task board. You stay in Blender. The AI comes to you. This matters more than it sounds. Artists context-switch constantly. Removing one more trip to an external chat app is not a small thing for anyone doing production work. Installing It Takes About Two Minutes The process is standard Blender addon installation: 1. Download blender-ai-mcp-addon.zip from the GitHub Releases page https://github.com/anujasid/blender-ai-mcp-addon 2. In Blender, open Edit → Preferences → Get Extensions → Install from Disk 3. Select the ZIP, then enable the addon in the list 4. Open the 3D Viewport and press N to open the side panel 5. Open the MCP Pro tab, pick a provider, paste your API key, choose a model, and start chatting Supported providers: OpenAI/ChatGPT, Anthropic Claude, Google Gemini, Ollama local , OpenCode Zen, and any custom endpoint that follows the OpenAI, Anthropic, or Ollama API format. The Ollama support is worth calling out specifically — studios that want inference on their own hardware, offline, with no per-token cost, can do that here. What the 21 Scene Tools Actually Do The addon ships 21 tools across four categories. This is a deliberately practical scope: - Inspection: list all objects, read scene structure, query materials and properties - Creation: add cubes, spheres, cylinders, planes, lights, cameras - Edit: move, duplicate, delete, parent objects, assign materials, apply modifiers - Utility: take viewport screenshots, run arbitrary Python code, look up Blender API documentation, export scenes Practical prompts that work well: - “Create a 3-point studio lighting setup for the selected mesh” - “Make three material variations for the cube: matte white, brushed metal, translucent glass” - “Rename all objects in the scene using their mesh type as prefix” - “Take a viewport screenshot so I can see the current state” These are not glamorous tasks. They are the repetitive ones. This is where AI in a 3D pipeline earns its place — not by generating a character from a text prompt, but by handling the mechanical work you already know how to do and hate doing manually. Multi-Agent Team Workflows: The Feature Everyone Is Talking About For complex tasks, the addon lets you configure a Team chain. You define a manager agent, worker agents, and a reviewer, each with its own model, role, and set of instructions. A simple example: a planner agent reads the scene and writes a brief, a worker agent executes bpy Python calls, and a reviewer agent takes a viewport screenshot and validates the output against the brief. This is the first Blender addon that surfaces multi-agent coordination to artists without requiring them to write any orchestration code. Whether it scales beyond small scenes and simple tasks is still an open question — the project is four days old — but the architecture is there. What It Cannot Do Read This Part Blender MCP of any flavor does not generate meshes. The Blender Foundation confirmed this in May 2026: no generative AI feature exists or is planned in Blender's codebase. The AI manipulates geometry that already exists, via Python. Organic shapes, sculpt mode, and complex rigging setups remain fully manual work. Complex multi-step operations also drift. The more steps between instruction and result, the more likely something lands subtly wrong. Generated geometry node graphs and custom shader setups are brittle across Blender versions. The model cannot see your render — it reads scene data. If you want to generate a character from a text prompt, you want mcp-for-blender https://github.com/ahujasid/mcp-for-blender with its Meshy integration. If you want a pipeline assistant that handles repetitive scene work and stays in your viewport, blender-ai-mcp-addon is the cleaner option right now. For an honest breakdown of what current Blender MCP tools can and cannot do in production, the 3D AI Studio comparison https://www.3daistudio.com/3d-generator-ai-comparison-alternatives-guide/blender-mcp is worth reading before you commit. How It Compares to the Alternatives | Tool | UI location | Providers | Mesh gen | Stars | |---|---|---|---|---| | blender-ai-mcp-addon | Inside Blender | All + Ollama | No | ~2K 4 days | | mcp-for-blender ahujasid | External client | Most | Meshy | 34.6K | | dcc-mcp-blender | Embedded server | External clients | No | 63 | | Blender Lab MCP | External client | External clients | No | Official | For a full comparison including setup complexity and pipeline fit, the Blender MCP server comparison on Stray Spark Studio https://www.strayspark.studio/blog/official-blender-mcp-server-comparison-2026 covers all current options. Who Should Install This Now Install it now if you are a technical artist who already uses Blender for scene assembly, lighting, and material work, and you want to offload the mechanical parts of that work. Install it if your studio wants to test local model inference in a 3D pipeline without spending on API tokens. Install it if you are a developer who wants to explore multi-agent workflows in a 3D context — the architecture is interesting even at this early stage. Wait if you need production stability. The project is four days old. The multi-agent features in particular should be treated as experimental. File issues, contribute fixes, and check back in a month.