DEV Community

Cover image for I discovered I was losing money on my own users and had no idea
Ghiles Asmani
Ghiles Asmani

Posted on

I discovered I was losing money on my own users and had no idea

#ai #llm #saas #buildinpublic

Six weeks ago I got my OpenAI bill and something felt off.

I dug into it and found one user had cost me $40 that month. They were on a $9 plan. I had been losing money on this person for weeks with no idea.

I asked other founders and they all had the same thing going on. Everyone was just hoping the math worked out until the invoice arrived. Nobody had a real answer for tracking this.

So I built Weckr.

Two lines of code wraps your existing OpenAI, Anthropic or Gemini client. From that point you can see exactly which users are profitable and which ones are eating into your margins in real time. It also detects agent loops and fires a Slack alert if a single user session burns through 50K tokens in 5 minutes. That one feature alone can save you thousands if you run any kind of agentic workflow.

Here is what the integration looks like:

import { Weckr } from '@weckr/sdk'

const wk = new Weckr({
  apiKey: 'wk_xxx',
  plans: { free: 0, pro: 29 }
})

const result = await wk.chat(openai, {
  model: 'gpt-4o-mini',
  messages: [...],
  userId: user.id,
  feature: 'ai-summary',
  plan: user.plan
})
Enter fullscreen mode Exit fullscreen mode

Same LLM call. Weckr logs cost and margin per user in the background with zero added latency.

Built the whole thing solo in 6 weeks. TypeScript and Python SDKs on npm and PyPI.

Curious if anyone else has hit this problem. How are you tracking AI costs per user right now?

Try it at useweckr.com or see the demo without signing up at https://useweckr.com/demo

Top comments (0)