Run LAMMPS in the cloud.
An agent can create the simulation. Point Hermes, Claude, Cursor, or Grok at the MCP server (MCP, the Model Context Protocol, lets an AI assistant use tools). It writes the input, submits the job, and reads the log. The same project is in the browser. There is no cluster account.
| Agents | api.quickmdsim.com/mcp. Token or OAuth. |
|---|---|
| Binary | LAMMPS 10 Sep 2025 (feature release), GPL-2.0. Packages: MOLECULE, COLLOID, KSPACE, MANYBODY, DIPOLE. Not compiled: ReaxFF, MEAM, machine-learning potentials. |
| Compute | Free plan: 1 CPU core. Basic and Pro plans: 2, 4, 8, or 16 cores. The Pro plan can also use 32 or 64 cores, or one NVIDIA L4 GPU (GPU guide). Small systems (under about 2,000 atoms) do not run faster on more cores. |
| Billing | You pay in CPU-hours. One CPU-hour is one core for one hour. A job costs (cores reserved) × (minutes it runs). One second on the L4 GPU costs 12 CPU-seconds; GPU jobs stop after 60 minutes. New accounts get 0.5 CPU-hours free after email verification. No card needed. Included hours on paid plans are listed in the app. |
| Scope | We run your input as written; we do not check the physics. Good for small and medium jobs. For very large, long runs, a supercomputer allocation is still the better choice. |