What is this?
This is a Blender add-on that turns the Blender viewport into a chat window with an AI agent (OpenCode): describe in text or show an image of what you want, and the agent generates Python (bpy) code that the add-on runs directly in your scene.
This is a Blender add-on that turns the Blender viewport into a chat window with an AI agent (OpenCode): describe in text or show an image of what you want, and the agent generates Python (bpy) code that the add-on runs directly in your scene.Is this a standalone app or a Blender plugin?It is a Blender add-on (plugin) installed from inside Blender via Preferences, not a separate application — the agent chat lives in a sidebar tab inside the 3D viewport.
It is a Blender add-on (plugin) installed from inside Blender via Preferences, not a separate application — the agent chat lives in a sidebar tab inside the 3D viewport.How do I install and start using it?Download the zip from Releases and install it via Edit > Preferences > Add-ons > Install, open the sidebar (N) → AI Copilot tab, click 'Install OpenCode' to fetch the runtime (~80MB), set your provider/API key/model in AI Settings, then click Start Server and send your first message.
Download the zip from Releases and install it via Edit > Preferences > Add-ons > Install, open the sidebar (N) → AI Copilot tab, click 'Install OpenCode' to fetch the runtime (~80MB), set your provider/API key/model in AI Settings, then click Start Server and send your first message.Do I need to install OpenCode myself, or get an API key first?No separate OpenCode install is needed — the add-on downloads a portable OpenCode runtime into its own folder the first time you click 'Install OpenCode'. You do need an API key for your chosen cloud provider (OpenAI, OpenRouter, Anthropic, Groq, DeepSeek), unless you use a local Ollama model, which needs no key.
No separate OpenCode install is needed — the add-on downloads a portable OpenCode runtime into its own folder the first time you click 'Install OpenCode'. You do need an API key for your chosen cloud provider (OpenAI, OpenRouter, Anthropic, Groq, DeepSeek), unless you use a local Ollama model, which needs no key.What's the catch with letting the agent run code automatically?Generated bpy code executes automatically in your scene (Auto-run), and if it errors the agent reads the server log and tries to self-fix it; for safety the agent is explicitly denied shell access and can only edit files inside its own working folder.
Generated bpy code executes automatically in your scene (Auto-run), and if it errors the agent reads the server log and tries to self-fix it; for safety the agent is explicitly denied shell access and can only edit files inside its own working folder.Which operating systems does it support?Windows, macOS and Linux, on both x64 and arm64 — the installer automatically fetches the matching OpenCode build for your platform from GitHub Releases.
Windows, macOS and Linux, on both x64 and arm64 — the installer automatically fetches the matching OpenCode build for your platform from GitHub Releases.Can I create or edit a model from an image instead of just text?Yes, you can attach an image file or a viewport screenshot and the agent generates bpy code based on it, but this requires a vision-capable model such as gpt-4o, a Claude model, or llava.
Yes, you can attach an image file or a viewport screenshot and the agent generates bpy code based on it, but this requires a vision-capable model such as gpt-4o, a Claude model, or llava.Can I run everything fully locally without any cloud API?Yes, you can set the provider to local Ollama (e.g. the llava model), which needs no API key at all, though image input still requires a vision-capable model.
Yes, you can set the provider to local Ollama (e.g. the llava model), which needs no API key at all, though image input still requires a vision-capable model.What's planned next for the add-on?The roadmap includes a cancel button for running agent sessions, exporting agent-generated meshes/assets directly into the .blend file, and an image-to-3D mode powered by Meshy or Tripo.
The roadmap includes a cancel button for running agent sessions, exporting agent-generated meshes/assets directly into the .blend file, and an image-to-3D mode powered by Meshy or Tripo.What are similar projects?genex-desktop, mcp-for-blender, NextChat, quivr, routeVSCODE, AstrBot, opensquilla, devika
genex-desktop, mcp-for-blender, NextChat, quivr, routeVSCODE, AstrBot, opensquilla, devikaWhich topics is it in?groq, 3d-graphics, blender, openrouter, code-generation, gpt, chatgpt, 3d, ollama, deepseek, generative-ai, anthropic
groq, 3d-graphics, blender, openrouter, code-generation, gpt, chatgpt, 3d, ollama, deepseek, generative-ai, anthropic