Myriox sends a rushed commuter, a wheelchair user, a lost delivery worker, and dozens more AI personas walking through your plan. Each one reasons about why a layout fails — then a compliance auditor checks every corridor they struggled with against real building codes.
The old way
That's exactly the problem — paper doesn't walk through the building.
A static 2D floor plan can't tell you that the corridor outside the elevator becomes a chokepoint at 5pm, or that the accessible route dead-ends at a fire door.
Legacy crowd tools like AnyLogic or MassMotion model people as particles obeying flow equations — they don't capture why a lost delivery driver circles back, or why a parent avoids a crowded stairwell.
Accessibility gaps, compliance violations, and layout dead-ends usually get caught in a walk-through review — after the budget for changes is already gone.
How it works
Paint walls, doors, ramps, stairs, and service points on the grid editor, or drop in a reference image. Mark your entry and exit points.
Choose how many of each persona to deploy — rushed commuter, wheelchair user, parent with a stroller, delivery worker, elderly visitor — each with its own goals and mobility profile.
LangGraph drives each agent through an Azure OpenAI reasoning call every tick. A GridValidator resolves collisions in real time as dots move across your plan.
Every stall, dead-end, or narrow corridor is checked against real ADA/IFC clauses via retrieval-augmented search — with the exact code cited, not a vague warning.
No black box — here is exactly what each simulation does.
Every agent is powered by an LLM with a distinct objective, mobility profile, and memory — it reasons about why a layout fails, not just that it's congested.
LangGraph orchestrates concurrent agents navigating your grid, causing emergent bottlenecks the same way real crowds do — a poorly placed exit becomes a visible traffic jam.
A RAG-backed Compliance Auditor checks every corridor, ramp, and doorway against ADA and building-code clauses as agents encounter them — no separate audit pass required.
Catch the accessible-route gap or the delivery dead-end in a simulation, not during a change order after the walls are up.
Beyond a single walkthrough
A single simulation tells you a layout works on an average day. Myriox is built to push past that — testing the edges before you commit to a layout.
Scale a run up to 100 agents across all five personas to find the crowd density at which a corridor, exit, or checkout queue tips over into a genuine bottleneck — not just a busy afternoon.
Cast rushed commuters alongside a wheelchair user, a parent with a stroller, and a lost delivery worker in the same run — the friction between different mobility profiles sharing one corridor is exactly what a physics-only sim can't surface.
Deterministic geometry checks (narrow corridors, missing ramps) run alongside an LLM that reasons over vector-retrieved ADA/IFC clauses — so a finding is either a hard rule violation or an explained judgment call, never a vague warning.
Why Myriox
Legacy crowd simulators (AnyLogic, MassMotion) and generative-design tools (Forma) solve adjacent problems well. Neither one simulates a person who thinks.
Set up your first grid, cast five personas, and get a cited compliance report in minutes — no install, no CAD license required.