{"slug": "archbench-control-blender-through-chat", "title": "ArchBench – Control Blender through chat", "summary": "ArchBench launched as a desktop AI assistant that controls Blender through chat, letting users describe a modeling change, review the proposed action and affected objects, then apply it in the native application with supported Undo. The tool's first integration was demonstrated on macOS with live scene context, and broader release testing is underway; free accounts start with GPT-5.6 Luna while GPT-6 Astra is the paid default reasoning model. Public MCP connectors on GitHub are planned starting with Blender, and ArchBench's IFC modeling and reformatting workflows for architecture, engineering and construction remain in design.", "body_md": "### Describe what needs to change.\n\nStart from live scene context. Describe a blockout, an arrangement, or a change to your selection. Free accounts start with GPT-5.6 Luna; GPT-6 Astra is the paid default.\n\nYOUR AI ASSISTANT FOR 3D DESIGN\n\nA desktop companion for your design software. Describe an idea, review the proposed changes, and keep creating in your model. Starting with Blender.\n\nYour design tools. An assistant alongside.\n\nYOUR SOFTWARE. A NEW WAY TO WORK.\n\nStay in the tools you know. Choose an application to see where its connector stands.\n\nOur first integration. Live scene context, reviewed modeling actions, and native Undo have been demonstrated on macOS. Broader release testing is underway.\n\nPublic MCP connectors on GitHub are planned, starting with Blender. Each will document its supported versions and workflows.\n\nBUILT AROUND THE MODELING WORK\n\nStart with a clear instruction. Keep the project context close. Produce a result you can keep working with.\n\nPlans, material schedules, and specifications are next. Planned IFC workflows will connect project information to building elements, with units and properties checked.\n\nReview the proposed operation and affected objects. ArchBench applies supported changes in the native application and reads back the result.\n\nFROM INTENT TO EDITABLE WORK\n\nA desktop workflow that keeps you close to the design, from the first request to the final decision.\n\nPick an enabled AI model and connect a supported design application.\n\nStart with the live model. File and design-data inputs are coming next.\n\nAsk in your own words. Inspect the proposed action and the objects it affects.\n\nApprove the action, inspect the native result, and use supported Undo when needed.\n\nAI WORKFLOWS FOR ARCHITECTURE, ENGINEERING & CONSTRUCTION\n\nWe’re designing ArchBench to turn conversations into structured modeling, revision, and data preparation workflows. Our mission graphs guide the task; IFC gives building information a shared structure; native connectors carry out supported changes.\n\nCarry the brief, decisions, and progress through a defined workflow. Resolve missing inputs, review the proposal, and continue from the right step when the conversation resumes.\n\nWork with building elements, properties, units, and relationships. Design modeling and reformatting tasks around explicit schema mappings and checks.\n\nConnect reviewed changes to supported operations in your design software. Read back the result and retain the evidence needed for the next step.\n\nThe workflow we’re designing\n\nAstra provides general reasoning and tool-calling capabilities. ArchBench’s intended value is the AEC workflow system around the model: mission state, building-data mappings, native operations, and result checks, brought together for everyday project work. GPT-6 Astra is the default reasoning model within that system.\n\nNo. IFC provides a common schema for building information. Each workflow still needs a supported schema version, explicit source-to-target mappings, and checks for missing or unsupported information. Native editing history and application-specific features may not transfer. ArchBench’s IFC modeling and reformatting workflows are in design. [About the IFC schema](https://technical.buildingsmart.org/standards/ifc/)\n\nAEC harness architecture in development. Mission graphs exist in our chat platform; their use for desktop AEC tasks and the IFC layer are planned. GPT-6 Astra is the default model; live availability depends on account and release access.\n\nGPT-6 ASTRA · ARCHBENCH\n\nGPT-6 Astra is the default. ArchBench brings your project context, modeling tools, and change review into the conversation. Live access depends on your account and release.\n\nYOU STAY IN THE DESIGNER’S SEAT\n\nWorking with a real project means knowing what will change, where it will happen, and how to get back.\n\nArchBench\nInspect the proposed action and affected objects before you approve.\n\nResults stay in the connected application as native model work.\n\nNative Undo and recovery are qualified for each supported workflow.\n\nArchBench is a desktop AI assistant for 3D modeling and building design. It brings live model context into a conversation so you can describe and review changes in your design software. Blender is the first integration under development; guided AEC workflows and IFC building-data support are planned.\n\nFree accounts start with GPT-5.6 Luna. GPT-6 Astra is ArchBench’s default model for paid accounts. You can choose another eligible model independently of your design software. Live access depends on your account and the current release. IFC-aware AEC mission workflows are still in development.\n\nBlender is the first integration under development. SketchUp, Rhino/Grasshopper, Revit, and Archicad are planned, with more connectors to follow. Support will be released for specific workflows and application versions.\n\nPublic MCP connectors are planned, starting with Blender. The aim is to let compatible AI assistants connect to supported modeling software, with setup instructions and a clear list of supported actions and versions. Public repositories and installation links will appear here when published.\n\nFile-backed commands are on the roadmap, beginning with selected images, text briefs, and bounded PDF workflows. Structured schedules and connected design data follow. The current Blender foundation uses live scene and viewport context.\n\nStart with Download to see the packaged desktop app for your operating system. The downloads page shows which releases are ready and which are still being prepared. Once installed, sign in and connect Blender; AI usage depends on your account and release access.\n\nLET’S BUILD YOUR NEXT WORKFLOW\n\nSign in to save your place. Share your link to earn priority. We’ll email you when your access is ready.\n\nCookies help us improve ArchBench. Accept to allow analytics and advertising cookies (Google, Meta), or decline to keep them off. [Privacy](/legal/privacy-policy)", "url": "https://wpnews.pro/news/archbench-control-blender-through-chat", "canonical_source": "https://archbench.com", "published_at": "2026-09-10 05:37:17+00:00", "updated_at": "2026-09-10 05:53:14.635760+00:00", "lang": "en", "topics": ["ai-products", "ai-tools", "generative-ai", "ai-agents", "developer-tools"], "entities": ["ArchBench", "Blender", "GPT-5.6 Luna", "GPT-6 Astra", "GitHub", "IFC", "macOS"], "alternates": {"html": "https://wpnews.pro/news/archbench-control-blender-through-chat", "markdown": "https://wpnews.pro/news/archbench-control-blender-through-chat.md", "text": "https://wpnews.pro/news/archbench-control-blender-through-chat.txt", "jsonld": "https://wpnews.pro/news/archbench-control-blender-through-chat.jsonld"}}