Run multiphysics simulations

Set up structural, thermal, electromagnetic, acoustic, fluid and coupled studies on your CAD by answering a few questions. RunMat meshes the part, solves fast on GPUs, CPUs and clusters, and checks each result against your limits.

Download RunMat

Read the guide

RunMat's guided simulation setup on a wall-hook bracket: the user picks a strength study, gives S355 steel, clicks the three mounting holes and the hook seat on the model, and runs it; the solved displacement shows the hook seat deflecting 0.017 mm, 0.083 mm under the 0.1 mm limit, with the held holes at zero.
0:00 / 0:00

Guided setup on a steel wall hook: answer a few questions, click the mounting holes and the hook seat, and run the solve.

Set up a study by answering questions

Open a part and say what you need to know, such as how far it bends under load or how hot it runs. Guided setup asks how the part is held, loaded and made, lets you click the faces it means, and flags a missing support, terminal or inlet before anything runs.

It writes the study as it goes, checks it and solves it, and the answer comes back as a number with the result on the part. The file it leaves behind is short enough to review line by line.

Read the guideHow RunMat Agent works

cantilever.step is a 1 m steel beam. Fix one end, push 1 kN down on the other, and check the tip deflection stays under 1 mm.

Writing and solving cantilever.fea

The tip deflects 0.200 mm, well inside the 1 mm limit and matching the 0.200 mm beam theory gives.

Displacement magnitude from the saved analysis study

Mesh and solve your own CAD files

Import a STEP, IGES or BREP file and name the faces and bodies your study refers to. RunMat meshes the exact geometry with its own mesher and saves a mesh quality report next to it. The same file and settings always give the same mesh.

After the solve, open the result in RunMat Desktop or in a browser and turn it around, or read it from a script and compare it with your own hand calculation.

Geometry and CAD importStudies and solves

Von Mises stress from the saved analysis study

Von Mises stress in the same cantilever with the load in pascals (100 kPa on the end face, 1 kN in total), solved by RunMat. The tip deflects 0.200 mm; beam theory gives 0.200 mm.

Every study is a short text file you can read

Describe the part and what you need to know. An AI model writes the study: the geometry, materials, loads and analyses, each in a few lines with units. RunMat lists every input back to you, marked as something you stated, something it derived, or a proposal you still need to confirm. The model's changes arrive as a checked edit you accept or reject.

In most simulation tools that setup is hundreds of choices in menus and dialogs, saved in the tool's own project file. A RunMat study is a short text file, so you can read it line by line, see exactly what changed between two versions and copy it to the next part. The same study always gives the same mesh and the same result.

RunMat checks units and regions before it meshes anything, so a mistake, yours or the model's, is caught in about a second with the line and the reason.

cantilever.fea
geometry beam {
file = "./geometry/cantilever.step"
units = mm
region body : solid { select name "Beam" }
region fixed_end : face { select name "Fixed End" }
region loaded_end : face { select name "Loaded End" }
}
material steel {
elastic {
youngs_modulus = 200 GPa
poisson_ratio = 0.3
}
}
model loaded_beam {
assign material.steel to geometry.beam.region.body
constraint fixed_end {
structural.fixed { target = geometry.beam.region.fixed_end }
}
load end_traction {
traction on geometry.beam.region.loaded_end = [0, 0, -100000] N
}
}
study cantilever_static {
geometry = geometry.beam
mesh = mesh.beam
model = model.loaded_beam
analysis static { structural static {} }
}
terminal
$ runmat check cantilever.fea
error[fea.semantic.model]: traction must be force per area, found N
--> cantilever.fea:22:55
|
22 | traction on geometry.beam.region.loaded_end = [0, 0, -100000] N
| ^^^^^^^^^^^^^^^^^ expected Pa
checked cantilever.fea: 1 error(s), 0 warning(s)

Check stress, deflection and temperature against your limits

Add an assessment to the study: the quantity to check, such as peak von Mises stress or tip deflection, and the limit, such as 250 MPa or 0.25 mm. After the solve, each assessment comes back passed or failed, with the governing value and where it occurs. In the cantilever above, the deflection check passes at 0.200 mm against 0.25 mm.

When the result can't support a decision, for example because the error estimate straddles the limit or a peak was only sampled at a few points, the assessment comes back indeterminate and says why. Change a limit later and RunMat checks the saved result again without re-solving.

For rules that don't fit a single limit, such as a margin of safety from your own method, write the check as a MATLAB function and import it. Every result also lists the mesh, solver and settings that produced it, and the analyst and the design engineer can open the same 3D result at once and see the same verdict.

AssessmentsHow results are recorded

cantilever.fea
output peak_stress = max(
analysis.static.field.element.von_mises_stress,
over = geometry.beam.region.body,
)
output tip_deflection = max(
magnitude(analysis.static.field.node.displacement),
over = geometry.beam.region.loaded_end,
)
assessment stress {
value = output.peak_stress
maximum = 250 MPa
evidence = authored_samples
}
assessment deflection {
value = output.tip_deflection
maximum = 0.25 mm
evidence = authored_samples
}

Sweep designs and chain analyses

Declare the inputs to vary and their values, such as a service load of 0.5, 1.0 and 1.5 kN or ten wall thicknesses. Each case is a full solve, the cases run in parallel, and the outputs you name come back in the order you listed the values. If a case fails, the rest still finish and the failed one is listed with its error.

Chain analyses in one study: solve the temperature field, apply it as a preload, then compute the natural frequencies of the heated part. RunMat checks each hand-off before anything runs, and the analyses share one mesh with no export and re-import between them.

Run a parameter sweepCoupled physics

bracket.fea
sweep load_envelope over study.bracket_response {
axis service_load {
half = 0.5 kN
nominal = 1.0 kN
proof = 1.5 kN
}
output peaks = collect(case.output.peak_stress.value)
}
heated_modes.fea
study heated_modes {
geometry = geometry.part
mesh = mesh.volume
model = model.structure
analysis heat { thermal steady {} retain heated }
analysis preload after analysis.heat {
thermomechanical preload { temperature = analysis.heat.state.heated }
}
analysis modes after analysis.preload {
structural prestressed_modal {
prestress = analysis.preload.state.geometric_stiffness
modes = 1
}
}
}

Use MATLAB functions for loads, geometry and checks

Some inputs are easier to compute than to type: a force pulse, a temperature profile from test data, a part shape built from a few parameters, a pass or fail rule. Write them as MATLAB functions and import them into the study. RunMat checks each function's arguments and units along with the rest of the study.

You can also run studies from a MATLAB script, change a parameter in a loop, and read the results back as arrays, next to the formula you would check them against.

Use MATLAB functions in studies

impactPulse.m
function force = impactPulse(time, peak, duration)
...
phase = min(time ./ duration, 1);
force = peak .* sin(pi .* phase) .* (time <= duration);
end
cantilever_impact.fea
import function impactPulse from "./impactPulse.m"
...
model struck_beam {
...
signal pulse : force over time {
generated by function.impactPulse {
arguments {
peak = 1 kN
duration = 6 ms
}
sample from 0 s to 50 ms at 4 kHz
interpolation = linear
}
}
excitation impact {
force {
target = geometry.beam.region.loaded_end
signal = signal.pulse
direction = global([0, 0, -1])
}
}
}

Solve on your GPU, every core or a cluster

RunMat solves a study on every core of one computer and on its GPU, and spreads large models and many sweep cases across machines: three computers per person on Professional, larger clusters under Enterprise.

Meshing and solving run locally with no internet connection, and the same study gives the same mesh and result on a laptop, a workstation or a server.

Scale to a cluster

LaptopRemote serverLARGE CLUSTER

Use RunMat with Claude Code, Codex or its built-in agent

Connect Claude Code, Codex or any MCP client, or use RunMat Agent. Describe the part, how it is held, the loads and the result you need; the agent writes the study, RunMat checks and solves it, and each change arrives as an edit to accept or reject.

AI integrationsAgent docs

FAQ

01Do I need to be a simulation expert?

No. Guided setup asks how the part is held, loaded and made, and writes the study with you. If you know finite element analysis, write and edit studies directly.

02Can I set up a study from Claude Code or Codex?

Yes. Connect Claude Code, Codex or any MCP client to RunMat Desktop, describe the part and the question, and the study is written, checked and run in RunMat, where you review it.

03Which CAD formats can I import?

STEP, IGES and BREP files import as exact geometry. STL and OBJ files can be opened for viewing.

04Can I read the solver's source?

Yes. The solver is part of the RunMat runtime, and its source code is available at https://github.com/runmat-org/runmat. The runtime is free for individuals, including freelancers; students and faculty at an academic institution, for teaching or research there; contributors to an open-source project, for work on that project; organizations with 250 or fewer people and US$10M or less in annual revenue; and anyone, for non-production use. Organizations with more than 250 employees and contractors or more than $10M in annual revenue use it under a commercial license.

RunMat license

05Can it run without an internet connection?

Yes. Meshing and solving run on your computer with no internet connection, and so does plain .m code. A network connection may be required for security and permission checks.

06How do I know the results are right?

Each physics family has a validation page with the reference cases its solvers are checked against, such as beam deflection and steady heat conduction. The cantilever on this page deflects 0.200 mm at the tip under 1 kN, against 0.200 mm from beam theory.

Verification and validation

Set up your first study

Download RunMat

Read the guide