Boids Flocking Simulator

What you just saw

Each agent looks only at its immediate neighbors, adjusts its velocity according to three rules — stay close, don't collide, match heading — and yet hundreds of agents organize into a coherent flock. No leader, no global plan, no central control. The murmurations you see in starling videos follow exactly this logic.

The science behind it

Craig Reynolds published the boids algorithm in 1987 while working on computer graphics for SIGGRAPH. The three rules — cohesion, separation, alignment — became the template for swarm behavior in film VFX (including Lord of the Rings battle scenes) and robotics. The model is also the canonical example of emergence: complex collective behavior from trivial local rules.

Try these experiments

1. 1. Classic flock

   Settings: cohesion=0.01 separation=0.05 alignment=0.05

   What to look for: Clean murmuration-style flocking.

2. 2. No cohesion

   Settings: cohesion=0

   What to look for: The flock disperses. Agents still avoid each other, but they no longer cluster.

3. 3. No separation

   Settings: separation=0

   What to look for: Agents collapse to a single point and stop moving usefully.

4. 4. Max alignment

   Settings: alignment=0.3

   What to look for: All agents fly nearly parallel. Rigid, crystalline flock.

from scirouter import SciRouter
client = SciRouter(api_key="sk-sci-...")
result = client.simulation.run(
    model="boids",
    params={"num_agents": 100_000, "cohesion": 0.01,
            "separation": 0.05, "alignment": 0.05},
    steps=1000, seed=42,
)