CFD without the
text editor.
A desktop CFD workbench. Import geometry, mesh it, define the physics, run the solver and post-process the results — without ever opening a dictionary or a terminal.
No account, no email wall. The free tier runs every module and every analysis type in them — 250,000 cells of fluids, 100,000 nodes of solid.
26
Analysis types
Each with its own solver and fields
16
Turbulence models
Plus five viscosity models
v2606
Solver core
Bundled — nothing to install
107
Solvers bundled
The full OpenFOAM catalogue, shipped
A case, start to finish
Five stages, in one application, on your machine. No handoff between a mesher, a solver and a viewer that each want the geometry in a different form.
01
Import geometry
STL or STEP. Check, clean and heal it on the way in, so a bad surface is a message rather than a mesh that quietly leaks.
02
Mesh
A background block, refinement where the gradients are, and layers on the walls that matter. The cell count is shown before it runs.
03
Set up physics
Choose the analysis type and the solver follows from it, with the fields, dictionaries and boundary conditions it actually requires.
04
Solve
On every core you own, or on a machine that is not the one you are sitting at. Residuals plot live, so a case that is not converging says so early.
05
Post-process
Slices, contours, streamlines and probes in the viewport. Forces and coefficients as numbers you can put straight in a report.
The physics, in every tier
Twenty-six analysis types across five families, with a turbulence model chosen on top of whichever you pick. None of it is held back from the free build — what the tiers change is scale, support and seats.
Flow
Incompressible through to supersonic, plus atmospheric, scalar transport and rotating frames. Steady where the physics has a steady answer, transient where it does not.
Heat
Convective, conjugate and pure conduction. Fluid and solid solved together, with heat crossing between them.
Multiphase
A sharp interface with VoF, interpenetrating phases with Euler-Euler, plus cavitation, sediment, shallow water and waves.
Reacting and particles
Diffusion flames with chemistry, premixed fronts, sprays, and discrete particles tracked through the flow.
Beyond fluids
Solid stress, electrostatics, magnetostatics and magnetohydrodynamics, on the same mesh machinery as everything else.
Turbulence
Not an analysis type but a layer on top of one: RANS, LES and DES families, every model available on every tier including the free build.
The actual competitor is a terminal
The solver is free. Your time is not.
Most people evaluating this already drive an open-source solver from a command line and pay nothing for it. So the question is not whether a solver is worth money — it is whether hand-writing the mesh dictionary, the numerical schemes and thirty boundary-condition entries per case is worth the hours it takes.
Every dictionary this application writes is one you did not have to get right from memory, and every field it knows a boundary condition needs is one you did not discover was missing forty minutes into a run.
The commercial comparison is simFlow at around €8,000 a year for comparable desktop coverage, and SimScale, which is capable but cloud-only and metered.
This is £550 a month, on your own machine, with your cases on your own disk. There is no feature gating between the paid tiers — the difference is seats and support, not capability.
End to end
Geometry in, report out.
Six steps, all of them inside one window. The batch flags drive the same path from a script.
Import the geometry
Every polygon format, plus STEP, IGES and BREP through OCCT. Or start from a built-in worked example and import nothing at all.
Mesh it
snappyHexMesh with refinement regions, boundary layers and quality controls — set up in a panel, not in a dictionary.
Define the physics
Sixteen turbulence models, Newtonian and four non-Newtonian viscosity models, per-patch boundary conditions that know which fields they need.
Run it
Locally or over SSH, with the log streamed back and residuals, monitors and problems in a dock while it solves.
Post-process
Surfaces, slices, contours, streamlines, vector glyphs, thresholds and oil-flow lines, with probes and plot-over-line. No ParaView round trip.
Report it
An A4 PDF with results in engineering units, the split of drag into pressure and friction, run cost, and a verification checklist that states the threshold behind each verdict.
Coverage
Twenty-six analysis types.
Each one derives its own solver, field set, dictionaries and boundary conditions. Every one was checked by generating a case and putting it through the bundled solver — not by reading the dictionaries it wrote.
- Incompressible
- Compressible
- Supersonic (density-based)
- Solid conduction
- Convective heat transfer
- Conjugate heat transfer
- VoF and compressible VoF
- Cavitation
- Miscible mixing
- Sediment / drift flux
- Euler–Euler
- Marine and waves
- Potential free surface
- Shallow water
- Atmospheric boundary layer
- Lagrangian particles
- Spray
- Reacting and premixed combustion
- Passive scalar transport
- Rotating frame
- Solid stress
- Electrostatics
- Magnetostatics and MHD
Advanced concepts — MRF, AMI, overset, porous media and sources — are set up as cell zones that topoSet actually creates. Full capability list.
How do you know it is right
Measured, not asserted.
Two different questions, answered separately: does the application set the case up correctly, and does the answer match reality. The first is checked continuously across the whole catalogue. The second is what benchmarks are for.
Verification — every release
One hundred and seven solvers ship with the application, and the catalogue is exercised by generating each case, meshing it and running it rather than by assuming it still works. All twenty-six analysis types are put through the same route end to end.
It is a measurement, not a test suite that pins whatever the code did the day it was written. That distinction is the reason it keeps finding things.
Validation — against published data
The lid-driven cavity at Re = 100, checked against Ghia, Ghia & Shin (1982), Table I — the standard problem for an incompressible code. The interior agrees to about half a percent on a deliberately coarse 20 × 20 mesh.
Published in full, down to the per-station numbers, and reproducible from the application: Worked example, then Generate mesh and Run. Further benchmarks are run in development; they are written up as they are finished, and only appear here once they are.
Building that first benchmark immediately found two faults that had been in the software the whole time — every PISO case failed at start-up, and every sealed domain stopped without a pressure reference. Both are fixed. We say so because a CAE vendor that tells you what its own checks caught is more useful than one that implies nothing was ever wrong.
Try it on your own geometry.
The free build is the same binary as the paid one. Install it, mesh something real, and decide from there.