Four engineering files, and the report that came out of them.
Someone dropped four STEP files into a folder — a spur gear, a gearbox housing cover, a steering knuckle, a trim panel — and an AI assistant handed back a measured report on all of them. No CAD seat opened, no numbers typed by hand.
The gear's volume: 126.68 cm³. Surface area: 310.00 cm². The assistant computed them from the geometry itself. Here is how it got there.
It read each file through an ODA MCP Server, the piece that lets an assistant work with a CAD or BIM file instead of guessing at it. Reading gave it the model's structure: which parts exist, how they sit together, which views the original author defined.
For each part it computed volume, surface area and centre of mass, then estimated the mass for a range of candidate materials and proposed one that fit the part it recognised: a hardened steel for the gear, a cast aluminium for the thin-walled housing cover.
Then it drew. Every view already defined in the model was rendered from that view's own camera. The gear came back annotated from the model's own manufacturing data. The knuckle came back with callouts naming the tie-rod eye, the wheel hub bore, the caliper mounting and the kingpin bores.
Last it assembled the document: what it had done, an overview of the four models, then a section per part with its properties, a material table and its views. One pass, start to finish.
An assistant that summarises a CAD file gives you a description. One that measures it gives you something you can put in front of a customer, because the numbers came from the geometry.
