30 hours, countless bugs, five concurrent ClaudeCode instances and one of the coolest demos ever.

Today, ~11,000 satellites are actively operating in orbit. Within four years, that number is projected to exceed 100,000. But the number of operators monitoring them? Roughly the same. Satellite operations are still largely manual — critical decisions like anomaly response, collision avoidance, and intent analysis are made by small teams while orbital environments get denser by the month.

So we built something to change that.

OrbitalShield is a real-time space defence platform that fuses live orbital data from SpaceTrack, runs it through Bayesian threat scoring, and deploys a multi-agent AI pipeline to detect, analyse, and recommend responses to hostile satellite behaviour.

In our demo, a Chinese satellite executes an unannounced manoeuvre toward a US reconnaissance asset. OrbitalShield flags the orbit change, pulls 730 days of TLE history, cross-references five intelligence databases, classifies the threat — and executes a trajectory manoeuvre before a human operator would have finished reading the alert. All rendered on a 3D globe with live trajectories, threat lines, and an AI console showing the agents' reasoning in real time.