Tips for Implementing Prompt Caching with OpenAI and Clay API

Anybody here successfully using prompt caching with the OpenAI or Anthropic APIs inside Clay’s API column? I’m trying to cut costs. Right now the entire email is generated by Anthropic and ends up at roughly $0.04–$0.05 per email, even though about 90 percent of the prompt never changes. I’ve tried setting up prompt caching, but I’m not seeing any savings. If someone has this working, how did you set it up?