Blender-AI-MCP-Suite README

项目的 README,来自 github.com/findavoulot1973/Blender-AI-MCP-Suite。合并为一个 markdown 文件:README.md。

Blender AI MCP Suite

Blender AI MCP Suite is an AI assistant and built-in Model Context Protocol (MCP) server for Blender that turns plain-language briefs into inspected, approved and reversible 3D scene operations — with a visual Studio workspace, multi-agent teams, 21+ scene tools and your choice of AI models.

License Python Windows MCP Blender AI Release

Download

Picture the workflow: you type "Build a night product stage with a pedestal, a camera and three lights" — the assistant inspects the scene, plans ordered steps, places the pedestal, aims the camera, sets the lights, applies a matte blue material and returns a viewport screenshot for review. Every step is approved in chat, every change is reversible through a named undo checkpoint, and the scene is export-ready when you are.

Why

3D work is full of mechanical steps — arranging props, tuning materials, re-exporting variations — while generic chat tabs cannot see your scene and blind scripts edit without review. Blender AI MCP Suite puts the assistant inside Blender: it inspects the scene first, then builds, edits and exports through a validated tool set you approve step by step, with undo checkpoints on every mutation. You iterate at the speed of conversation and keep full creative control.

Without / With

Without With
Hours of manual clicks for layouts, materials and exports One plain-language brief becomes a finished, checked scene
Chat tabs that cannot see your 3D scene An assistant that inspects objects, transforms and materials before acting
Blind scripts and hard-to-trace changes Every operation shown in chat, approved and reversible
One model stretched across every kind of task A team of manager, workers and reviewers, each fit for its role
Asset hunting across many websites Poly Haven, Sketchfab and Poly Pizza searched from one chat
Lock-in to a single AI vendor Cloud providers, local Ollama models or custom endpoints

Quick Start

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

Features

How it works

Blender AI MCP Suite runs inside Blender as one extension with three cooperating parts: the AI Suite sidebar (chat, scene outline, viewport snapshots, approvals and the task board), the connection manager (saved providers, keys and models, kept separate for work and personal use) and the built-in MCP server (an authenticated local Streamable HTTP endpoint for external clients). A brief is inspected through 21+ scene tools, split into ordered steps that run one by one — each visible, approved and reversible — while larger jobs are delegated to a manager, workers and reviewers; external clients such as OpenCode drive the same tool set through the MCP endpoint.

Usage

Configuration

Setting What it controls Default
Provider AI connection: OpenAI, Anthropic Claude, Google Gemini, OpenRouter, OpenCode Zen, Ollama or a custom API Set on first run
Model Tool-calling model for chat, vision and agents Chosen from the provider catalog
Saved connections Separate work and personal keys, multiple keys per provider Personal
Team chain Roles, instructions and models for manager, workers and reviewers Single assistant
MCP endpoint Authenticated local Streamable HTTP server for external clients Active with Studio
Interface language English, German, French, Spanish or Chinese English
Asset services Keys for Poly Haven, Sketchfab and Poly Pizza Empty
Undo checkpoints Named undo steps before mutating operations Enabled

Troubleshooting / FAQ

The add-on does not appear after installation. Confirm you are running Blender 4.2 or later on Windows x64, reinstall via Edit → Preferences → Get Extensions → menu → Install from Disk and enable Blender AI MCP Suite.

Windows blocked the downloaded ZIP. Right-click the archive, open Properties, check Unblock and confirm, then install it again.

The assistant does not answer or changes nothing. Check the saved connection details and key, select a model with tool-calling support, and approve the pending operation in the chat.

An external MCP client cannot connect. Keep Blender open while using Studio or external clients, confirm the client supports authenticated Streamable HTTP MCP, and copy the endpoint exactly as shown in the sidebar.

A team task stopped. A blocked task gets one correction attempt; a second failed review stops the workflow — open the task board, read the findings and resume from the last approved step.

Where do credentials live? In your user settings on your own machine; chat history lasts for the current session. Python tool validation is a guardrail, not a security sandbox.

Contributing

Spot a bug or miss a tool? Open an issue with your Blender version, Windows build and a short reproduction. Pull requests that add scene tools, interface languages or provider support are welcome. If Blender AI MCP Suite saves you clicks, leave a ⭐ star — it helps other artists find the suite.

Keywords

blender ai mcp suite · blender ai mcp · ai for blender · blender ai assistant · blender mcp server · blender mcp · model context protocol · mcp integration · ai blender addon · blender extension · blender addon · blender plugin · blender copilot · blender automation · blender scripting · blender python · bpy api · blender api · geometry nodes · 3d modeling · 3d automation · 3d workflow · ai scene builder · scene generation · ai modeling · ai materials · viewport screenshot · blender agents · blender multi-agent · ai agents 3d · blender chatgpt · blender claude · blender openai · blender gemini · blender ollama · blender openrouter · opencode blender · local llm blender · chat in blender · ai chat blender · blender workflow · product display scene · asset search blender · poly haven blender · sketchfab blender · poly pizza blender · blender export automation · blender undo checkpoint · studio workspace blender · visual chat blender · ai tools for blender · blender studio window

License

Released under the MIT license.

Blender-AI-MCP-Suite (findavoulot1973/Blender-AI-MCP-Suite)

正在搜索数据库…

这是什么?Blender AI MCP Suite 是 Blender 的一款 AI 助手,并内置 MCP 服务器。你可以用自然语言描述场景,它会检查场景、规划有顺序的步骤,并在你批准后以可撤销的方式执行这些操作。
它是独立应用还是 Blender 插件?它是一个以扩展形式安装的 Blender 插件,包含三个部分:带有聊天、场景大纲和审批功能的侧边栏;用于管理 AI 服务提供商的连接管理器;以及供外部客户端使用的内置 MCP 服务器。
怎样让它运行起来?从 Releases 下载 mcp-for-blender-pro.zip。在 Blender 中依次进入 编辑 → 偏好设置 → 获取扩展 → 菜单 → 从磁盘安装,选中该 ZIP 文件,然后启用它。之后打开 3D 视图 → N → MCP Pro,选择服务提供商并挑选模型。
安装前需要准备什么?需要 Blender 4.2 或更高版本,以及一个受支持服务提供商的 API 密钥,或者一个本地 Ollama 模型。该模型必须支持工具调用。
Windows 上安装时有哪些需要注意的问题?Windows 可能会拦截下载的 ZIP 文件。右键点击压缩包,打开“属性”,勾选“解除锁定”,确认后重新安装即可。
能在 macOS 和 Linux 上用吗?README 中的安装说明仅针对搭载 Blender 4.2 或更高版本的 Windows x64 系统。Blender 本身可以在其他系统上运行,但文档中没有给出相应步骤,因此 macOS 和 Linux 上的使用情况尚未确认。
没有网络也能运行吗?只有在使用本地 Ollama 模型,或在本机运行的自定义端点时才可以。OpenAI、Claude、Gemini 和 OpenRouter 需要联网,在 Poly Haven、Sketchfab 和 Poly Pizza 上搜索素材同样需要联网。
我的 API 密钥存放在哪里?密钥和连接信息保存在你本机的用户设置中。聊天记录仅在当前会话期间保留。
这个助手在场景中实际能做什么?它会先检查对象、变换、材质和修改器,然后借助 21 个以上的场景工具添加基本体、灯光、摄像机,并进行父子级设置、复制对象和指定材质。每一步都会在聊天中显示,并需要你批准。
助手能检查自己的工作成果吗?可以,前提是使用支持视觉的模型。视口截图会返回供检查,每一次修改操作都会生成一个带名称的撤销检查点,你可以回退到任意检查点。
多智能体团队是如何协作的?管理者会把较大的任务拆分为带有验收标准的分配项,工作者执行场景操作,审核者则负责批准或将问题退回。受阻的任务会获得一次修正机会,任务看板会记录每一次尝试。
OpenCode 等外部工具能控制它吗?可以。内置 MCP 服务器是一个经过身份验证的本地 Streamable HTTP 端点,OpenCode 等外部客户端可以通过它使用同一套工具。使用时 Blender 必须保持打开。
大小
<0.1 MB
星标
498
时长
6天
提交 · 90 天