OpenFOAM external aerodynamics · MCP-native
AI engineering simulation that shows its work.
Describe a CFD problem in plain English. SimPrompt generates an editable OpenFOAM setup, runs the real solver, and returns a validated report. The engineer approves the schema before anything runs — and walks away with the case files.
Vanilla OpenFOAM case files you own·Matches your house template, shows the 12-line diff
What we simulate
External aerodynamics on OpenFOAM, across four sub-niches. Same SIS-CFD schema underneath, same case files in your hands at the end.
Vehicle aerodynamics
Ahmed-body and production-vehicle external aero. Drag, lift, and surface-pressure fields, exported as OpenFOAM case files.
Steady · incompressible · k-ωSST · single-phase
Pedestrian wind comfort
Urban-canyon wind at pedestrian height. Velocity contours and comfort criteria, ready for planning-permit reports.
Steady · incompressible · k-ωSST · single-phase
Wind-turbine blades
Blade-section external aero with boundary-layer resolution. Pressure-coefficient distributions for design iteration.
Steady · incompressible · k-ωSST · single-phase
Drone & UAV propellers
MRF rotating-frame boundary conditions are first-class. eVTOL, FPV, and small-UAV teams iterate on rotor geometry in days.
Steady · incompressible · k-ωSST · single-phase
How it works
Three steps. The AI is on either side of the solver — never inside it.
Describe in plain English
"Simulate flow over an Ahmed body at 25 m/s." Plus your house template (or one of ours). The Intent Agent generates an editable SIS-CFD schema — your controlDict, fvSchemes, and boundary conditions as YAML — anchored on your conventions.
Diff against your house template
Bring your own baseline system/ and 0/dictionaries. SimPrompt fills the deltas and surfaces a side-by-side diff — twelve lines you actually need to think about, not two hundred you don't. Approve, edit, or reject any line. Nothing runs until you do.
Real solver, real logs
OpenFOAM runs on AWS Batch. You see the actual solver output. The Evaluator Agent summarises results and flags convergence issues. Case files exportable, always — bundled with the schema and a reproducibility recipe so a colleague (or your client) re-runs it on a fresh box.
Why SimPrompt and not the alternatives
Other AI-CFD tools exist. SimPrompt's three durable differences:
Vanilla OpenFOAM you own
Real OpenFOAM. Not a proprietary fork, not a black-box surrogate. Take the case directory and run it anywhere simpleFoam runs — on-prem, on your AWS account, on a workstation under the desk. Hand it to a client, a regulator, or a supplier without re-licensing.
Anchored on your house template
Upload your team's baseline system/ and 0/ once. Every generated case starts there and shows you only the deltas. Your conventions for y+, domain extents, ref area, force coefficients — preserved, not overwritten by an AI's idea of “best practice.”
Compresses the unbillable hours
Setup, reruns, reports, hand-off packaging — the work nobody bills for. SimPrompt does the boilerplate; you keep the engineering judgment. A 4-day case turns into a 4-hour case, and the report writes itself from the case directory.
Benchmarks
We will not publish a number we cannot reproduce. Currently building.
| Benchmark | Reference | SimPrompt result | Status |
|---|---|---|---|
| Ahmed body — drag coefficient | Ahmed (1984) | — pending — | Targeting Week 4 |
| NACA0012 — lift & drag at α | NASA TMR | — pending — | Targeting Week 4 |
| NASA TMR turbulence cases | turbmodels.larc.nasa.gov | — pending — | Targeting Week 6 |
When we publish, we publish the case files, the schema YAML, and the reproducibility recipe. If you can't re-run our number, we don't ship the number.
Join the waitlist
We are looking for senior CFD practitioners as design partners. If you run OpenFOAM commercially, you're exactly who we want to talk to.