PyMCNP Documentation

PyMCNP is a Python package and CLI for interfacing with MCNP. PyMCNP parsers MCNP files, visualizes simulation geometries, and runs MCNP simulations in parallel.

Example

Inputs are Python objects, so loops can build geometries that are impractical to write by hand, and the results go straight into pandas and matplotlib. This builds 441 iron discs of random thickness over a rough concrete floor, tallies the neutron flux of a point source, and plots the result:

import numpy as np
import pymcnp
from pymcnp.inp import card
from pymcnp.inp.card import surface

rng = np.random.default_rng(20260101)
surfaces, cells = [], []
for ix in range(-10, 11):
    for iy in range(-10, 11):
        x, y, thick = ix * 10.0, iy * 10.0, float(rng.uniform(1, 10))
        disc = surface.Rcc(j=len(surfaces) + 1, vx=x, vy=y, vz=50.0, h1=0.0, h2=0.0, h3=thick, r=4.0)
        surfaces.append(disc)
        cells.append(card.Cell_0(j=len(cells) + 1, m=1, d=-7.874, geom=-disc, options=['imp:n=1']))
# ... air, ground, source and FMESH tally, then inp.to_file('ground.inp')

# After `pymcnp run`: one row per mesh voxel.
(tally,) = pymcnp.Meshtal.iter_tallies('meshtal')
frame = tally.to_dataframe()

Setup Flux

The full example includes the complete model, the geometry plot, and the flux plot.

Automating with AI Coding Agents

PyMCNP ships a skill file (SKILL.md) that describes the API, the parsing contract, and the common pitfalls to an AI coding agent such as Claude Code. With the skill installed (for Claude Code, copy it to ~/.claude/skills/pymcnp/SKILL.md), an agent can write, scan, run, and plot MCNP models for you from a short description, so automating MCNP becomes easy even for complex geometries. The example above was created this way from a short prompt. Always review generated models and results before relying on them.

Table of Contents