Chinilla

Chinilla turns static architecture diagrams into runnable systems. You drop seven universal building blocks onto a whiteboard-style canvas — Person, Step, Storage, Decision, Trigger, Tool, Channel — wire them with connections that carry latency and capacity, attach behaviors, and hit run. A discrete-event engine pushes packets through the design in real time so you can watch queues fill, retries fire, circuit breakers trip, and bottlenecks surface before you write any code. The behavior layer is where it earns its keep. Eight programmable modes — passthrough, filter, split, delay, retry, circuitbreaker, batch, replicate — attach per node, and queueing is universal: set capacity on any component and it becomes a queue. Backpressure kicks in at 80% queue capacity, FIFO overflow drops the oldest packet, and random behaviors use a fixed seed, so same design plus same inputs gives the same result every time. That reproducibility is what makes before/after comparisons worth doing. It's aimed at system design interview prep, architecture learning, whiteboarding new services, teaching, and research prototyping — the fast pass before you reach for AnyLogic or Arena. The free tier covers cloud projects, the full simulation engine, all components and behaviors, templates, stability scoring with Monte Carlo runs and parameter sweeps, timeline views, and PNG/SVG/Mermaid export. Pro adds live collaboration with cursors, interview mode with 8 hand-authored rubrics, an AI design assistant, code-to-diagram from GitHub URLs across 15+ languages plus YAML/JSON/TOML/XML, animated GIF export, and publishable live URLs with embeddable iframes. Positioning is straightforward: Excalidraw and diagrams.net draw pictures; Chinilla reasons about topology, throughput, capacity, and failure modes. It is not a load tester, deployment visualizer, or wire-level network simulator.