MCDC Agent

This agent is in testing. Generated scripts may require manual review. Please let Gunnar on slack know if you have any issues

An AI agent that guides you through building MC/DC neutron transport simulations. Uses Google Gemini to help you generate simulation input scripts. Currently does not support the various MCDC techniques like population_control, branchless_collision(), etc., they will be added in the future.

Setup

1. Install

Option A: User Install (Recommended) Install directly from GitHub without cloning:

pip install "git+https://github.com/CARRE-PSAAP/MCDC-agent.git"

Option B: Clone and Install

1. Clone the repository:

   git clone https://github.com/CARRE-PSAAP/MCDC-agent.git
   cd MCDC-agent

2. Install:

   pip install .

2. Configure Gemini API

You need a Google Gemini API key from Google AI Studio.

export GEMINI_API_KEY="your-api-key-here"

If you encounter rate limits, you can set up billing (it will cost a few cents) or switch models by setting the GEMINI_MODEL environment variable:

export GEMINI_MODEL="gemini-2.5-flash"  # or "gemini-2-flash"

Usage

The agent provides a CLI command mcdc-agent.

Interactive Mode

Best for beginners. The agent guides you step-by-step.

mcdc-agent interactive

Input Script Generation (Recommended)

Generate a script directly from a prompt.

mcdc-agent generate "[Simulation description]"

Options

• mcdc-agent generate --file prompt.txt -o run.py: Read from file and save output.
• mcdc-agent --help: Show all available commands.