Stigmery / FAQ

Frequently asked questions

Quick answers. Need more depth? The docs cover the same topics in detail.

What is Stigmery?

Stigmery is a free, browser-based tool for building agent-based models (ABM). You define small worlds where many simple agents follow simple rules, run the simulation, and watch the global patterns that emerge from local interactions. It runs entirely in your browser - nothing to install. An AI co-pilot can build or tweak the same model in plain English.

Is it free?

Yes, free to use. No sign-up required to open the bundled predator-prey example or to start building a model. Sign-in is only needed if you want to save and re-open your work across devices, or share a model with a public URL.

Do I need to know how to code?

No. You can describe what you want in plain English (or Dutch) and the AI co-pilot will edit the model for you. If you do want to write rules yourself, the built-in DSL is a tiny expression language designed to be readable by anyone who has seen a spreadsheet formula. There's also a "code escape hatch" for the rare case you need to drop into a TypeScript subset.

Is it like NetLogo?

Stigmery takes the turtles-and-patches paradigm that made NetLogo so effective in education and rebuilds it for the modern web, with first-class AI tooling. The mental model is similar: a 2D grid of patches, mobile agents on top, simple rules that fire each tick. It's not a port - you cannot load .nlogo files - but a NetLogo user will recognise the ideas immediately.

How does the AI work? Is it just ChatGPT?

The assistant uses OpenAI's chat-completion API under the hood, but it doesn't write free-form code. It calls a closed set of typed tools - add an agent type, set a property, write a rule, configure a plot - the same tools the user interface exposes. That means every AI change is recorded in the model's change log, attributed to "ai", and you always see what was modified. You and the AI share one source of truth.

What about my privacy and my models?

Models are private by default. They live in your account and only become reachable via a URL when you explicitly switch a model to "public". The site only collects lightweight, anonymised usage events (app opens, model loads, tab switches) and never stores chat messages or rule text on the server. Account email addresses are used only for the magic-link sign-in.

Where does the name come from?

Stigmery is a play on stigmergy, a term coined by Pierre-Paul Grassé in 1959 for the way social insects coordinate indirectly through traces left in a shared environment. Ants don't tell each other where to go - they leave pheromone trails, and other ants follow. That's exactly what this tool is about: complex global behaviour emerging from simple local rules.

Can I share a model?

Yes. Save a model, switch it to public in the library dialog, and copy the URL. Anyone with the link can open it, run it, and fork it into their own account.

What runs in the browser vs on the server?

The simulation, rendering, and editor all run in your browser. The only thing that happens server-side is the AI proxy: the browser sends a tool-calling request to the server, which forwards it to OpenAI with the API key kept private. Nothing about your model or chat history is persisted on the server beyond what you explicitly save.

Who is this for?

Primarily students, teachers, and curious tinkerers who want to play with agent-based modelling without installing a desktop tool or learning a programming language first. It is also useful as a quick sketchpad for researchers who already know ABM and want to prototype an idea faster.

Is there a Dutch version?

The interface is in English, but the AI co-pilot understands and answers in Dutch (or any language OpenAI's model handles). For now, the documentation is English-only.