Hi all,
I’m an AI student taking an agent-based modeling course, and I built a free,
open-source tool that connects NetLogo to AI assistants — and it has
first-class CoMSES Net integration, which is why I’m sharing it here.
What it does
NetLogo MCP is a Model Context Protocol
server. Once connected to an MCP client (Claude Desktop, Claude Code, Cursor,
VS Code, and others), an AI assistant can:
- Create NetLogo models from a plain-English description — with real
interface widgets (sliders, switches, buttons, monitors, live plots) - Run simulations and collect tick-by-tick reporter data
- Set up and run BehaviorSpace parameter sweeps headlessly
- Export the view as PNG, inspect world state, sample agents, read patch grids
The CoMSES Net part
The server can search the CoMSES model library, retrieve model metadata,
download and safely extract releases (with size caps and archive
validation), read ODD documentation and source files, and open NetLogo
models from the library directly. There’s also a built-in NetLogo 6->7
transition guide so AI assistants can help port older .nlogo models to 7.0.
A workflow I use in my coursework: “Search CoMSES for opinion dynamics
models, summarize the ODD of the top result, then open it and run a
baseline experiment.”
Details
- GitHub: GitHub - Razee4315/NetLogo-MCP: The first MCP (Model Context Protocol) server for NetLogo, enabling AI assistants like Claude to create, run, and analyze agent-based models through natural conversation. · GitHub
- Install:
pip install netlogo-mcp(requires NetLogo 7.0+ and Java JDK 11+) - MIT licensed, free forever; 25 tools; GUI and headless modes; CI-tested
I built this because nothing like it existed when I needed it for class.
I’d genuinely value feedback from this community — especially on what would
make it more useful for research reproducibility workflows (e.g., would
automatic ODD drafting from model code be useful? Better experiment
export formats?).
Thanks!
Saqlain