My co-founder and I built EchoCache, a semantic caching layer for LLM APIs (OpenAI, Anthropic, Gemini, or anything OpenAI-compatible). The problem: repeated user queries that are worded differently but mean the same thing still cost a full API call each time. Exact-match caching doesn't catch this. EchoCache embeds incoming prompts, checks similarity against past queries, and returns a cached response on a match — otherwise it passes through to your LLM provider as normal. It's a drop-in wrapper — point your app at EchoCache instead of directly at the provider, nothing else changes. Free tier for smaller traffic, paid plan for higher-volume use like production support agents. Site: https://echocache.vercel.app/ Happy to answer questions about the matching approach, false-positive handling, or anything else.

EchoCache

Stop paying twice for identical LLM queries

Stop paying twice for identical LLM queries