Tracking the cost for multiple API calls is a pain. So I Built this

Community
, , , ,
1

Since the Community category states “Topics should be related to what is happening in the news, sharing cool projects you are working on, and conversations around AI safety.” I thought I should share a project I built E2E here.

Pretty sure there’s people in here using tools such as n8n, Make, and/or Zapier to orchestrate workflows involving various AI tools and you might have scenarios calling OpenAI for text generation or even image/video generation AI tools as well.

But here’s the annoying thing. While those services simplify the automation, tracking the combined cost of all those different AI API calls, including OpenAI API calls, within your workflows is anything but simple. How much does that complex customer support automation really cost per run, factoring in every AI module it touches? I found myself manually writing scripts to log data or making rough guesses, lacking a clear, unified view.

I was frustrated. So I built AIBillingDashboard.

So how does it help me? This is what I noticed so far:

  1. Unified AI Cost Dashboard: Now I can see spending across all the AI services my workflows use (OpenAI, RunwayML, Eleven Labs, etc.) in one central place.
  2. Scenario Cost Insights: By logging usage data from workflows (e.g., using HTTP requests or tools like Google Sheets/Airtable to capture token counts/service calls) and uploading it, I can visualize the AI costs associated with specific scenarios.
  3. Optimize AI Spend: Now I can identify which AI modules within automations that are the most expensive. For example, could a different model or provider save me more money without sacrificing quality completions/outputs?
  4. Centralized Billing: Tracking billing cycles and getting payment alerts for all the disparate AI services that charge during different times of the month is pretty helpful.

So, I built this because managing the costs of my own, and others, multi-step AI automations was becoming unmanageable.

How are you tracking the AI costs generated within your workflows? Any clever workarounds or major challenges?