Synsci handles computational chemistry workflows end to end. Activate the chemistry mode and the agent treats your prompts the way a computational chemist would. It picks the right level of theory, the right software stack, and the right hardware for the structure or system in front of you.

Activate the mode

synsc web --mode chemistry
For a one-shot run:
synsc run --mode chemistry "Optimize this MOF structure with PBE-D3 and report the band gap"
To set chemistry as the default for your current working directory:
synsc mode set chemistry

Who this is for

  • Computational chemists running DFT, post-HF, MD, or QM/MM workflows
  • Medicinal chemists doing structure-based drug discovery
  • Materials chemists modeling MOFs, COFs, zeolites, and porous frameworks
  • Synthesis researchers using retrosynthesis and reaction prediction
  • Quantum chemists working with multireference methods and excited states

Frameworks the agent uses

When you describe a chemistry task, the agent reaches for these natively. You don’t need to invoke them by name.
CategoryStack
DFT and ab initioPySCF, ORCA, NWChem, Psi4
Molecular dynamicsGROMACS, OpenMM, LAMMPS, AMBER, NAMD
CheminformaticsRDKit, Open Babel
Drug discoveryAutoDock Vina, GNINA, DiffDock, equibind
Quantum chemistryCASSCF, CCSD(T), TDDFT, EOM-CC workflows
ML potentialsMACE, Allegro, NequIP for ML-driven dynamics
Free energyOpenFE, Yank, GROMACS FEP

Workflows

Each recipe shows what to type and what comes out.

DFT geometry optimization

Optimize this MOF structure with PBE-D3 in VASP. Use a 4x4x4 k-point grid and converge forces below 0.01 eV/A. Report the band gap when done.
The agent reads your structure, generates INCAR with PBE-D3 settings and the requested convergence, sets up KPOINTS, submits the job locally or on Modal/Lambda for larger systems, and reports the band gap.

Molecular dynamics simulation

Set up a GROMACS simulation of this protein-ligand complex. Use AMBER ff14SB plus GAFF2 for the ligand. Equilibrate 5 ns NVT then 5 ns NPT, then 100 ns production at 300 K.
The agent generates topology files, sets up the equilibration phases with restraints, configures the production run, submits to compute, and runs basic analysis (RMSD, RMSF, hydrogen bonds).

Drug docking and virtual screening

Dock this 50,000-compound library against the active site of 1abc.pdb using GNINA. Filter for predicted affinity better than -8 kcal/mol and write the top 100 to results/.
The agent retrieves the protein, identifies the binding site, prepares the receptor, runs GNINA across the library on Modal GPUs, and writes the ranked results.

Retrosynthesis planning

Plan retrosynthetic routes for ibuprofen using a transformer-based model. Rank by predicted yield and feasibility, and surface the top three routes with reaction conditions.
The agent generates retrosynthetic trees with multiple disconnection options, ranks them by yield and step count, and returns the top routes with reagents and conditions per step.

Excited state calculation

Compute the first 10 singlet and triplet excited states of this molecule with TDDFT (CAM-B3LYP / def2-TZVP) in PySCF. Report oscillator strengths and the dominant orbital transitions.
The agent loads the molecule, runs the ground state SCF, runs TDDFT with the requested functional and basis, and reports excitation energies, oscillator strengths, and dominant transition character.

Free energy perturbation

Run a relative binding free energy calculation between these two ligands against PDB 5abc using OpenFE. Use 12 lambda windows per leg and 5 ns per window.
The agent sets up the alchemical transformation, generates the perturbation network, runs the windows in parallel on Modal GPUs, analyzes with MBAR, and reports the relative ΔΔG with statistical uncertainty.