LoopGain uses control theory to reduce agent loop costs by 92.8%

LoopGain is an open-source controller that replaces fixed max_iteration caps with real-time convergence detection for AI agent loops. In benchmark testing, it reduced API spend from $27.05 to $1.94 and improved median wall-clock speed from 30.9s to 2.1s while maintaining judge win-rates up to 0.63.