MOOSE MCP servers and AI agent skills

How AI connects to MOOSE

An MCP server or skill can only do what MOOSE lets programs do. MOOSE exposes 2 automation interfaces:

• HIT input files and pyhit (Python): MOOSE applications read input files in the hierarchical HIT format; MOOSE's first example input has blocks for Mesh, Variables, Kernels, BCs, Executioner and Outputs. The pyhit package in the MOOSE repository reads, edits and writes these files from Python using the same parser MOOSE uses, so a script or agent can change a parameter, read or edit comments and write the file back with pyhit's formatter. The VTB Assistant generates inputs in this format for BISON, Griffin and eight other MOOSE-based applications. Documentation
• MOOSE Python tools (Python): The MOOSE repository's python folder holds supporting packages, including the TestHarness, MooseDocs, moosetree, mooseutils, postprocessing tools and the Peacock GUI. MOOSE itself is LGPL-2.1 licensed and automatically parallel; its README cites runs on more than 100,000 CPU cores. Documentation

What INL offers

INL researchers describe MOOSEnger, a MOOSE agent that combines knowledge retrieval, HIT-aware parsing, language-server diagnostics and local or MCP-backed validation and execution (arXiv 2603.04756, revised 23 September 2026), and a follow-up on locally hosted models demonstrated on a Virtual Test Bed problem (arXiv 2608.15881, August 2026). No public MOOSEnger repository was found. INL's released AI tool is the VTB Assistant plugin.

Checked 4 Nov 2026. Sources: MOOSEnger: A Simulation-Aware AI Agent Framework for the MOOSE Ecosystem (arXiv 2603.04756), Deploying Frontier Agentic Technology in MOOSEnger (arXiv 2608.15881), vtb_assistant README (idaholab/virtual_test_bed)

What you can do with MOOSE and AI

• Find Virtual Test Bed models for a reactor concept, such as a molten salt or high-temperature gas reactor, filtered by the codes they use and whether they are transient.
• Draft a Griffin, BISON, SAM or Pronghorn input file starting from a VTB model that already shows the physics you need.
• Explain what each block of an existing VTB input file does, using the model's own documentation.
• Edit a parameter across a set of MOOSE input files with pyhit and write them back in a consistent format.

Taken from the listings' own documentation. Check each listing for exactly what it supports.

Installing

The listings here document setup for ChatGPT, Claude Code and claude.ai. Install steps, copied from each project's own documentation, are on each listing's page. Most MOOSE integrations run on your own computer next to MOOSE, so you need MOOSE installed and licensed on the same machine unless the listing says otherwise.