How to Track AI API Costs in Make (Integromat) Scenarios

TL;DR: Make's HTTP module lets you call any URL. Replace your OpenAI endpoint with https://tokonomics.ca/proxy/openai/chat/completions and your API key with a Tokonomics key. Every AI call gets tracked — cost per scenario, per model, per day. Setup takes 3 minutes.

Key Takeaways

• Make shows operations count, not AI cost — you don't know which scenario is burning $420/month in tokens
• Organizations waste 28% of cloud spend on average due to lack of visibility (Flexera, 2024)
• Use the HTTP module (not the built-in OpenAI module) to route through the proxy — 3-minute setup
• Tag per scenario and per client for agencies managing multiple accounts

Why Do Make Users Need AI Cost Tracking?

Make (formerly Integromat) is one of the most popular no-code automation platforms, and AI modules are everywhere — content generation, lead scoring, document processing, customer support.

The problem: Make shows you operations count, not AI cost. You know your scenario ran 5,000 times. You don't know those runs cost $420 in GPT-4o tokens — or that one scenario with a 4,000-token system prompt accounts for 70% of your total AI spend.

Without per-scenario cost visibility, you're flying blind. Most Make users discover their AI costs when the OpenAI invoice arrives, and by then, the money is already gone. Flexera's 2024 State of the Cloud Report found that organizations waste an average of 28% of their cloud spend due to lack of visibility, and AI inference costs follow the same pattern.

How Does the HTTP Module Connect to Tokonomics?

Instead of using Make's built-in OpenAI module (which doesn't support custom base URLs), use the HTTP module to call the Tokonomics proxy. The proxy forwards your request to OpenAI, records the tokens and cost, and returns the response unchanged.

Before:  Make → api.openai.com → response
After:   Make → tokonomics.ca/proxy/openai → api.openai.com → response

Your scenario logic stays the same. The AI response is identical. But now every call is tracked with cost, tokens, model, and latency.

Step-by-Step Setup

1. Create an HTTP Module

In your Make scenario:

1. Add a new module → search for HTTP → select Make a request
2. Set URL to: https://tokonomics.ca/proxy/openai/chat/completions
3. Set Method to POST

2. Configure Headers

Add these headers:

3. Set the Request Body

Select Body type: Raw, Content type: JSON

{
  "model": "gpt-4o-mini",
  "messages": [
    {
      "role": "system",
      "content": "You are a helpful assistant."
    },
    {
      "role": "user",
      "content": "{{1.input_text}}"
    }
  ],
  "max_tokens": 500
}

Replace {{1.input_text}} with the actual Make variable from a previous module.

4. Parse the Response

Select Parse response: Yes

The response JSON includes choices[0].message.content — map this to the next module in your scenario.

How Do You Use Claude (Anthropic) in Make?

For Claude models, use the same HTTP module pattern:

• URL: https://tokonomics.ca/proxy/anthropic/messages
• Headers: same Authorization: Bearer mk_... + Content-Type: application/json
• Body:

{
  "model": "claude-sonnet-4-6",
  "max_tokens": 1024,
  "messages": [
    {
      "role": "user",
      "content": "{{1.input_text}}"
    }
  ]
}

The same works for DeepSeek (/proxy/deepseek/chat/completions), Gemini, Mistral, and all supported providers.

How Do You Track Cost Per Scenario?

The key to useful cost tracking is tagging each scenario. Add the X-Feature-Name header with the scenario name:

X-Feature-Name: lead-enrichment

For agencies managing multiple clients in Make:

X-Metering-Tags: {"scenario":"lead-enrichment","client":"acme-corp","env":"production"}

These headers are stripped before reaching OpenAI. In the Tokonomics dashboard, you can filter and group costs by scenario name, client, or any custom tag.

This is how agencies answer the question: "How much did client X's AI automations cost this month?" McKinsey (2024) reports that 65% of organizations now use generative AI regularly, and per-client attribution is becoming essential for agencies billing AI services.

What Do You See After Connecting?

Your Tokonomics dashboard shows:

• Total AI spend across all Make scenarios
• Cost per scenario (via feature tags) — which automation is burning cash?
• Model breakdown — GPT-4o vs GPT-4o-mini vs Claude usage
• Daily trend — is spend growing? Stable? Spiking?
• Budget alerts — email or Slack notification at 70% and 90% of budget

For a detailed guide on setting up alerts, see how to configure budget alerts.

How Can Make Users Optimize AI Costs?

Once you have visibility, here are the most common savings:

1. Switch to cheaper models where possible

Most classification, extraction, and simple summarization tasks work fine with GPT-4o-mini ($0.15/1M input) instead of GPT-4o ($2.50/1M input). That's a 94% cost reduction for the same accuracy on simple tasks. OpenAI's own documentation (2024) recommends GPT-4o-mini for "straightforward tasks" where full GPT-4o reasoning isn't needed.

Use the cheapest LLM for each use case guide to match tasks to models.

2. Trim your system prompts

Every token in your system prompt is billed on every call. A 3,000-token system prompt costs $0.0075 per GPT-4o call. At 5,000 calls/month, that's $37.50 just for the system prompt. Cut it to 1,000 tokens and save $25/month from one scenario.

3. Set max_tokens appropriately

If your scenario only needs a one-sentence answer, set max_tokens: 100 instead of the default 4,096. You pay for output tokens — limiting them caps your cost per call.

4. Watch for zombie scenarios

Scenarios that are "paused" in your mind but still active in Make. Check for scenarios running on a schedule that no longer serve a purpose.

How Do Make Operations Compare to AI Costs?

Make charges per operation. A single AI call is one operation. But the Make cost and the AI cost are completely different:

For most users, AI tokens cost 3-15x more than Make operations. Yet Make shows operations, not token cost. That's the gap Tokonomics fills. According to a16z's GenAI cost analysis (2024), inference costs represent the single largest variable expense for companies building on LLMs, often exceeding compute and platform costs combined.

Frequently Asked Questions

Why not use Make's built-in OpenAI module?

Make's OpenAI module doesn't support custom base URLs. The HTTP module does, which is what makes the proxy integration possible. The HTTP module also gives you more control over headers (for tagging) and request structure.

Does the proxy slow down my scenarios?

The proxy adds approximately 30ms per call (benchmark data). For AI calls that take 500ms-3,000ms, this is invisible. Your scenarios won't notice the difference.

Can I track costs across Make + other platforms?

Yes. If you also use n8n, Zapier, or custom code, all calls through the Tokonomics proxy appear in the same dashboard. One unified view of AI costs across all platforms.

Get Started

1. Create a free Tokonomics account (100 calls/month free)
2. Copy your API key
3. Replace Make's OpenAI module with an HTTP module pointing to the proxy
4. Add X-Feature-Name headers for per-scenario tracking
5. Check your dashboard — your AI costs are now visible

All sources retrieved June 2026. Pricing: GPT-4o at $2.50/1M input tokens (OpenAI Pricing), GPT-4o-mini at $0.15/1M input tokens.