# Use Your ChatGPT Subscription in Make.com Scenarios

Call a Codex-backed OpenAI endpoint from Make.com with the HTTP module: exact module settings, response mapping, and what an AI-heavy scenario costs both ways.

*Published 2026-06-12 · https://proxyllm.ai/blog/use-chatgpt-subscription-with-make*

Make.com can run its AI steps on your ChatGPT subscription, with one change: instead of the native OpenAI app and a metered API key, use an HTTP module pointed at a Codex-backed endpoint. The request is the same OpenAI `chat/completions` call; the billing moves from per-token to your flat plan. Configuration takes one module.

Below: the exact module settings, the response-mapping gotcha specific to Make, and a worked scenario cost.

## Why there is no ChatGPT login for Make

ChatGPT plans and the OpenAI API are separate products, so Make's OpenAI connection wants an API key, not your ChatGPT password. The bridge is Codex: OpenAI's coding agent is included in ChatGPT plans and runs programmatically through `codex exec`, its documented non-interactive mode. [Codex Hosted](/blog/codex-hosted-setup-guide) runs the official CLI in an isolated container signed in by you (you run the login command on your own machine), and exposes one OpenAI-compatible endpoint. Intended functionality, with OpenAI keeping the final call; the deeper reasoning lives in the [n8n version of this guide](/blog/use-chatgpt-subscription-with-n8n), which covers the same split for that platform.

Make's native OpenAI app is hardwired for api.openai.com; unless your connection type exposes a base URL override, it cannot point elsewhere. The HTTP module can point anywhere, which makes it the dependable path.

## The HTTP module, field by field

Add **HTTP → Make a request** wherever the AI step belongs in your scenario:

```text
URL:             https://api.proxyllm.ai/v1/chat/completions
Method:          POST
Headers:         Authorization: Bearer YOUR_PROXYLLM_KEY
                 Content-Type: application/json
Body type:       Raw
Content type:    JSON (application/json)
Parse response:  Yes
Request content:
{
  "model": "gpt-5",
  "messages": [
    { "role": "system", "content": "You rewrite article summaries in our house style." },
    { "role": "user", "content": "{{1.summary}}" }
  ]
}
```

Drop scenario data into the JSON with Make's usual mapping panel, the `{{1.summary}}` above being whatever your trigger module outputs. Keep **Parse response** on; it turns the JSON reply into mappable fields instead of a text blob.

## Mapping the response

One Make-specific gotcha: Make indexes arrays from 1, not 0. The reply text therefore lives at:

```text
{{2.data.choices[1].message.content}}
```

where `2` is your HTTP module's number. Map that into the next module (a Google Doc, an email, a database row) and the scenario reads exactly like one built on the native app.

Two honest caveats. The Codex lane returns complete responses rather than streams, which suits Make perfectly, since a scenario module always waits for the full reply anyway. And keep embeddings or fine-tune calls on a real OpenAI API key; the model surface here is what Codex serves, meaning chat models.

## What a scenario costs, both ways

Worked example: a content pipeline that watches RSS, summarizes each item, rewrites it in house style, and runs a QA pass. Three model calls per run, 400 runs a day, averaging 2,000 input and 600 output tokens per call on GPT-5.4 (OpenAI June 2026 list: $2.50/M input, $15/M output).

| Monthly figure        | Amount                                         |
| --------------------- | ---------------------------------------------- |
| Model calls           | 400 × 3 × 30 = 36,000                          |
| Input tokens          | 72M × $2.50/M = $180                           |
| Output tokens         | 21.6M × $15/M = $324                           |
| **API total**         | **$504/mo**                                    |
| **Subscription path** | **ChatGPT Plus $20 + ProxyLLM $129/mo** |

A Plus window covers roughly $700 of API-equivalent work on our planning estimates, so this scenario fits with headroom; estimates, not guarantees, and the request log shows your real consumption per lane. Make bills operations identically on both paths, so only the OpenAI line moves.

Make bills operations and OpenAI bills tokens, and only one of those two meters has a flat alternative.

The honest flip side: below about $150 a month of OpenAI spend the token savings do not yet cover the flat fee, so at that level the reason to switch is building without the meter rather than a lower bill. Our Starter tier is $0 if you want the logs and dashboard first, BYO key, no markup, and the Codex lane is one module edit away when you want it.

The same pattern works in [Zapier via its webhooks step](/blog/use-chatgpt-subscription-with-zapier), and the [Make integration page](/integrations/make) keeps the copy-paste config. To price your own scenarios, the [calculator](/calculator) maps your current OpenAI bill to a plan tier in thirty seconds.