opensourceprojects.dev

A broadsheet for software that doesn't ask for your email

An MCP server and AI chat workspace inside Blender
GitHub RepoViews62

Project Description

View on GitHub

An AI Assistant That Actually Understands Your Blender Scene

You're deep in a Blender project, juggling objects, materials, and export settings, and you keep context-switching to a chat window to ask an AI for help—except that AI has no idea what's on your screen. It can't see your scene, can't touch your objects, and can't do anything except talk. The mcp-for-blender-pro addon is built to close that gap: it's an AI assistant that lives inside Blender, inspects your scene before it acts, and carries out actual operations while you watch.

What It Does

mcp-for-blender-pro is a Blender addon (Blender 4.2+) that combines a visual chat workspace with a built-in Model Context Protocol (MCP) server. The assistant can inspect your scene, build objects, edit materials, find assets, and run multi-step workflows—all from a companion Studio window that shows chat, scene objects, viewport snapshots, approvals, and a task board alongside your work.

Under the hood, it ships with 21 scene tools covering creation, inspection, editing, capture, asset search, Blender API consultation, and export. You connect it to the AI of your choice—OpenAI, Anthropic Claude, Google Gemini, OpenRouter, OpenCode Zen, Ollama, or custom endpoints speaking OpenAI Chat Completions, Anthropic Messages, or Ollama protocols. The built-in MCP server lets clients that support authenticated Streamable HTTP MCP (like OpenCode) connect too. Menus are available in English, German, French, Spanish, and Chinese.

Why It's Cool

The assistant gets context before it acts. This is the detail that matters most. Scene and object inspection means the AI isn't guessing what's in your file—it looks first, then operates. That single design decision is the difference between a chatbot and a collaborator.

You can run one assistant or a whole team. For quick changes, a single model in conversation is fine. For bigger jobs, you configure a Team chain: a manager that breaks requests into assignments and acceptance criteria, workers that perform scene operations, and reviewers that approve results or send back specific findings. Tasks run sequentially so scene edits stay ordered, and a blocked task gets exactly one correction attempt before the workflow stops. The task board exposes assignments, attempts, and review findings as things progress—which is a surprisingly honest way to handle AI failure modes.

  • Saved connections keep work and personal lives separate. Multiple keys for the same provider, distinct connection profiles—small thing, but it shows the tool was built by someone who actually uses it.
  • Local models are first-class. Ollama support means you're not locked into cloud providers if you'd rather keep things on your machine.
  • Image-capable models can read your viewport. If the model you pick supports images, it can interpret viewport screenshots—useful for visual feedback loops.
  • A real MCP server, not an afterthought. Connecting external clients through authenticated Streamable HTTP MCP opens the door to tooling beyond the addon itself.

One honest caveat from the README: you'll want models with tool-calling support, since that's how the assistant interacts with the scene. That narrows your model choices somewhat, but it's a reasonable constraint given what the tool does.

How to Try It

  1. Download mcp-for-blender-pro.zip from the latest release.
  2. In Blender, open Edit → Preferences → Get Extensions → menu → Install from Disk.
  3. Select the ZIP and enable the addon.
  4. Open the Studio workspace, add a connection (cloud provider, local Ollama, or a custom endpoint), load the model catalog, and pick a model.

From there, start with something small—ask the assistant to inspect your scene, or have it create a product-display layout and arrange objects. The quickstart guide and connections and agents docs cover the setup details, and the project is MIT-licensed.

Final Thoughts

This is a tool for artists who want faster iteration, technical artists automating repetitive work, and developers wiring AI into Blender pipelines. The team-chain workflow is the most interesting part—it acknowledges that AI makes mistakes and builds review into the process rather than pretending otherwise. If you already use AI in your creative work, this brings it into the viewport where it can actually help. If you're skeptical of AI in 3D workflows, the scene-inspection-first approach and local model support might be enough to make you curious.


Follow @githubprojects for more developer tools and open source projects.

Back to Projects
Project ID: blender-ai-mcp-addonLast updated: November 16, 2026 at 02:46 AM