At IMTS 2026 in Chicago (Sept. 15), SimScale launched an Engineering AI Agent inside PTC’s Onshape—delivered as an App Store app—so designers can ask a performance question in one prompt and get a reasoned simulation loop without exporting the model or waiting on a specialist. The agent sits in front of SimScale’s full cloud physics stack (CFD, FEA, thermal, electromagnetic), not a stripped “CAD-lite” checker, and reports results back in the same Onshape workspace where the geometry lives.
That matters because the CAD-to-CAE handoff has always taxed creativity: concepts stall while files bounce between departments, and only the “safe” ideas get validated. SimScale’s pitch is the opposite of a rough embedded solver. From one prompt—“will this bracket hold,” “how hot does this enclosure get,” “where is the pressure drop”—the agent reads active CAD (including geometry not yet cleaned for simulation), proposes physics, mesh, and boundary conditions, provisions compute, runs the job, and explains the answer in plain engineering language. The engineer can interrupt, override assumptions, or push variants at any step.
Governance is part of the product story. Runs go through the SimScale Agent API so actions stay auditable under the customer’s existing data controls, and the underlying project remains one click into the full SimScale UI for sharing or deeper specialist work. SimScale cites more than 900,000 platform users and a State of Engineering AI 2026 survey finding that nearly 90% of companies are putting AI governance in place as agentic workflows move into production. A follow-up Sept. 22 blog frames single validation loops as the starting point; longer Onshape branching/versioning workflows could let the agent explore parameter spaces against Physics AI predictions before spending a full validation run.
The industrial contrast is sharp. Traditional CAD-embedded simulation often stops at a quick check; anything off-menu forces a restart in a dedicated CAE tool. Here the agent works inside a normal SimScale project with the same validation-grade solvers production teams already trust. Early customers such as Withings have used earlier SimScale moves to compress design-to-prototype cycles; Convion’s fuel-cell work with Physics AI and Onshape is cited as collapsing months of optimization into under an hour. The risk, of course, is renting the reasoning layer and the compute—and trusting an agent’s proposed BCs without review.
For N23D, the lesson is timing and ownership. Putting a reasoning CAE agent at the CAD decision point is the right product instinct—and so is auditable provenance. Shops still need simulation assumptions, meshes, and revisions that survive when a cloud seat or App Store entitlement changes. N23D ships as one native Rust binary: local-first CAD with vaulted PLM-ready simulation history you control, not a prompt that only works while the SaaS clock is ticking.