ZenLLM

eams using LLM APIs can see the bill, but they still can’t explain it.

That’s why I’m building ZenLLM.

The problem I kept seeing:

- engineering knows the models

- finance sees the invoice

- nobody can clearly say which workflow, retry pattern, routing choice, or

context behavior is actually driving cost

So ZenLLM is built to make AI cost visible at the workflow level before teams

start guessing at fixes.

Lately I’ve been focused on making the benchmark and onboarding flow actually

usable, not just technically correct.