Driving Fusion via AI using MCP server add-in (announcement)

[OceanHydroAU]

I built and tested a general-purpose MCP AddIn for Fusion which I suspect has great potential in future; it's a careful architecture which generically exposes all API internals to the AI, no limits, making it possible to help with anything and everything you might ever need.

If anyone's interested in collaborating on this, let me know!  As great as it is right now, to be super-powered, we will need a smart-RAG pipeline added as well, so the AI knows everything it needs as it works, to guarantee correctness.  Big task - if you're good with AI and semantic search/embeddings, let me know!!  (This AddIn is built with included access to a local embeddings-generation module, semantic search, and a mini LLM - so all the hard stuff is done; feeding that brain is what's left to do!)

https://github.com/AuraFriday/Fusion-360-MCP-Server

https://www.youtube.com/watch?v=0T2XU4tzudQ

I also built a physics simulation engine into a python module, in preparation for adding this into Fusion as well (Fusion AddIns are python), and made that into an AI MCP tool as well...

https://pypi.org/project/fluidx3d/2.16.3/

https://github.com/gitcnd/FluidX3D-2.16-python
https://github.com/ProjectPhysX/FluidX3D/discussions/314

which should open up the door to letting you "see your design working" in real-time, on your screen, as you make stuff (and/or let the AI do that for you to...)