# blender-ai-copilot-bridge **An MCP bridge for Blender that exposes 27 supervised scene tools to OpenCode and any compatible client—authenticated Streamable HTTP on a local endpoint, with the visual chat studio and agent teams on the inside.** [download-url]: https://github.com/benzgasfuncring197110/blender-ai-copilot-bridge/releases [![License](https://img.shields.io/badge/license-MIT-blue)] ![Language](https://img.shields.io/badge/language-Python-3776AB?logo=python&logoColor=white) ![Platform](https://img.shields.io/badge/platform-Windows%20x64-0078D6?logo=windows&logoColor=white) ![Blender](https://img.shields.io/badge/Blender-4.2%2B-E87D0D?logo=blender&logoColor=white) ![MCP](https://img.shields.io/badge/Model_Context_Protocol-MCP-5E5CE6) ![Release](https://img.shields.io/badge/release-1.0.0-98D85B) ![Endpoint](https://img.shields.io/badge/endpoint-127.0.0.1%3A9876%2Fmcp-5E5CE6) [**Download**][download-url] Your coding agent gets a Blender hand: authenticate once, and the scene answers to `get_scene_info`, `add_cube` and `export_scene` like the rest of your toolchain. ## Why Your agents can read repos, write code and run tests—until the task says "now make it in Blender". blender-ai-copilot-bridge gives every MCP client a safe door into your scene: a local, authenticated Streamable HTTP endpoint that exposes the full supervised toolset—inspect, create, edit, capture, script, export—while approvals and undo stay in Blender. The bundled OpenCode launcher wires the connection for you; other compatible clients plug into the same endpoint with your token. ## Without / With | Without | With | |---|---| | Agents that stop at "open Blender and..." | A local endpoint they can call directly | | Screen-scraping and screenshot hacks | Clean tool calls with structured results | | Granting tools your whole machine | One authenticated endpoint, scoped by its token | | No record of what an outside tool did | Every operation visible and undoable in Blender | | Hand-copying scene data into prompts | `get_scene_info` feeds real context to the agent | | Working with one client only | Any client that speaks MCP over Streamable HTTP | ## Quick Start Every command below is self-contained: open PowerShell, copy it, paste it, press Enter. 1. Download **`mcp-for-blender-pro.zip`** from [Releases](../../releases/latest). 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 - Expose 27 supervised scene tools to any compatible MCP client. - Authenticate every request—a generated Bearer token, kept in the client's secret store or environment. - Launch OpenCode preconfigured—the bundled Portable launcher sets the Blender connection up for you. - Inspect scenes for your agents—object lists, structure and details returned as structured context. - Capture the viewport—screenshots that image-capable agents can actually see. - Keep control in Blender—approvals, cancellation and named undo apply to external calls too. - Export results—`export_scene` hands the finished scene to your pipeline. ## How it works blender-ai-copilot-bridge runs inside Blender as an extension: the MCP server publishes the supervised toolset on a local Streamable HTTP endpoint (`http://127.0.0.1:9876/mcp`) that authorized clients call with a Bearer token, and every call lands as a visible, undoable operation in Blender. Inside, the same toolset powers the visual Studio chat and optional manager–worker–reviewer agent chains; OpenCode Portable gets its connection configured automatically at launch. ## Usage **OpenCode in one click.** Launch **OpenCode Portable** from the sidebar, sign in to your providers, pick a model—and ask it to build in Blender; the bridge connection is already wired. **A build brief from your coding agent.** Point any MCP client at the endpoint with your token, then let it run `get_scene_info`, create the layout, set materials and finish with `export_scene`. **Context, not screenshots.** Feed real scene structure to your agents through the inspection tools instead of pasting viewport grabs into chat. ## Configuration | Setting | Default | What it does | |---|---|---| | MCP endpoint | `http://127.0.0.1:9876/mcp` | Local Streamable HTTP endpoint; enable the server in extension preferences and adjust the port if needed | | Access token | Generated when enabled | Bearer token for the endpoint; keep it in the client's secret store or environment | | Client examples | OpenCode and compatible apps | Any client that supports authenticated Streamable HTTP MCP | | Provider | OpenAI | OpenAI, Anthropic Claude, Google Gemini, OpenRouter, OpenCode Zen, Ollama or Custom API | | Model | Provider catalog | Pick a tool-calling model; enter the model ID manually when needed | | Workflow mode | Single assistant | Plain chat or a Team chain with one manager, workers and reviewers | | Approvals | Client policy + Blender controls | External clients manage their own approval policies; Blender keeps undo available | | Interface language | English | English, German, French, Spanish or Chinese | | Asset services | Optional keys | Service keys for asset providers that require them | ## Troubleshooting / FAQ **What is the endpoint address?** `http://127.0.0.1:9876/mcp` by default—enable the MCP server in the extension preferences first and keep Blender open. **How does authentication work?** Every request carries an `Authorization: Bearer` header with your access token; treat it like a password. **Can clients other than OpenCode connect?** Yes—any client that supports authenticated Streamable HTTP MCP; configure the endpoint and token in that client. **Is my scene exposed to the network?** The endpoint is local; the token gates every call, and operations remain visible and undoable in Blender. **Do approvals apply to external calls?** External clients manage their own approval policies—Blender-side undo and operation visibility still apply. ## Contributing Using the bridge with a client we should document? Open an issue with your setup and the call flow, or send a pull request. If blender-ai-copilot-bridge connects your agents to scenes, a star helps others find it. ## Keywords blender-ai-copilot-bridge · blender · blender 3d · blender addon · blender plugin · blender extension · blender scripting · blender automation · blender mcp · blender ai · blender assistant · blender copilot · blender chatgpt · blender llm · blender agents · bpy · 3d modeling · 3d automation · mcp · mcp server · model context protocol · ai for blender · mcp integration · streamable http · authenticated endpoint · access token · local server · opencode · external clients · coding agents · agent workflow · scene tools · scene inspection · structured context · viewport screenshot · execute code · export scene · bpy api · blender python · scripting · studio workspace · task board · agent teams · multi agent · model catalog · connection profiles · custom api · ollama · openai · anthropic claude · google gemini · openrouter · opencode zen · 3d automation · 3d workflow · windows 3d ## License Released under the [MIT](./LICENSE) license.