How it works

  • Once saved, your preference is appended after your prompt on every call to that model
  • No length limit — injected content is billed together with your input tokens
  • Platform quality & compliance guards always apply last and cannot be overridden