CacheCatch

CacheCatch helps developers optimize their token usage and minimize costs by auditing prompt-cache efficiency across various platforms. Key features: • Detects exact cache breakers in prompts • Estimates wasted spend due to inefficient caching • Provides specific prompt-layout fixes for improved efficiency • Works with LangSmith, Langfuse, Braintrust, and local IDE agents This open-source, local CLI tool analyzes your runs and sessions to generate comprehensive cache-loss reports. It highlights exact divergence points, calculates potential savings, and suggests immediate prompt adjustments to reduce token consumption significantly, potentially by up to 90%. CacheCatch ensures no prompts are uploaded, maintaining data privacy while offering deep insights. The tool delivers a fast, actionable report summarizing recoverable losses, identifying top leaking routes, pinpointing prefix breaks, and recommending the most impactful prompt layout changes. It supports cloud-traced sessions and direct IDE agent interactions, making it a versatile solution for any development team. Designed for engineers, CacheCatch provides the precise information needed to optimize spending and enhance the performance of their applications.