MolAgent covers the main branches of molecular and materials simulation and works with the codes research groups already use. Procedures are added and validated together with early-access users, and a code you rely on can be added.
Typical studies in each branch, and the codes the agent prepares, runs and analyses. Commercial codes such as Gaussian, VASP or Q-Chem run under your own licence.
Molecules and clusters with DFT and wavefunction methods.
Crystals, surfaces, interfaces and 2D materials with plane-wave and localised-basis DFT.
All-atom simulations of liquids, solutions, polymers, interfaces and biomolecules.
Binding, solvation and phase behaviour, and events too rare for plain molecular dynamics.
Larger systems and longer times, with groups of atoms represented as single beads.
Near-DFT accuracy within their training domain, at a small fraction of the cost, for screening and long simulations.
Parameters for new molecules and materials, checked against reference data.
The work around the calculations: structures in, numbers and figures out.
Five of these areas have an example study, with an interactive structure, on the Examples page.
Solvation structure, transport and stability of electrolytes; electrode surfaces and interphases.
Adsorption, reaction mechanisms and activity trends on metals, oxides and single-atom catalysts.
Defects, dopants, band alignment and interfaces in semiconductors and layered materials.
Binding, solubility, permeation and excipient interactions of drug molecules.
Curing, glass transition, mechanical response and interfaces in polymer materials.
Surfactants, membranes, emulsions and self-assembly, from all-atom to coarse-grained models.
Before it reports a result, the agent makes the checks an experienced practitioner would make.
Charge, multiplicity and stoichiometry; overlapping atoms; box size against interaction cut-offs; units.
SCF and geometry convergence; imaginary frequencies; spin contamination; sensitivity to the functional and basis set.
Cut-off and k-point convergence; slab thickness and vacuum; dipole corrections; magnetic states.
Equilibration of energy, density and box; energy drift; sampling length, block-averaged uncertainties and finite-size effects.
Overlap between neighbouring states; agreement between forward and reverse estimates; convergence with simulation time.
Walltime, memory and convergence failures are diagnosed, and the job restarts from the last good point with adjusted settings.
MolAgent runs on the machines and software stack you already have.
MolAgent is in alpha. We are installing it with a small number of research groups now; a private beta opens in 2027. Write to us with the systems you study, the codes you use and the machines you compute on.
Or write to contact@molagent.ai.