Skip to content

KazenAI documentation

KazenAI helps you stop AI agents from overspending before the provider call. Wrap an OpenAI or Anthropic client with monitor(), set a budget, and hard denials happen locally (and optionally against shared FinOps authority) before paid work runs.

This site is the public developer documentation for:

Product site: kazenai.com.

  1. Getting started
  2. Supported surfaces
  3. OpenAI guide or Anthropic guide
  4. Runnable Control demo
Phase What happens
Before provider dispatch Identity/attribution checks, pricing bound, local hold, optional shared reservation
Deny BudgetExceeded (or fail-closed unavailability) — zero provider calls
After a completed call Usage/cost settlement, optional soft KazenCircuitBreaker, event emission

See How a hard spending limit works and Reservation and settlement.