Burning through OpenAI API balance fast with the AI Agent node + GPT-5.5 — any way to use my ChatGPT Plus instead?

[ahmeqas]

My AI Agent node workflows are eating my OpenAI API balance way faster than
I expected — a handful of runs with GPT-5.5 and I’ve burned through more
than I budgeted for the month. Meanwhile I’m already paying for ChatGPT Plus
and barely using it.

Is there any way to point the AI Agent node (or a regular OpenAI node) at my
Plus subscription instead of metered API billing? I know the Agent node
expects an OpenAI-compatible credential, so in theory anything that speaks
that API should work — just not sure if anyone’s actually done this or if
I’m missing an obvious option in n8n itself.

Anyone found a way around this, or am I stuck paying for both?

[Anshul_Namdev]

Hi @ahmeqas Welcome!
ChatGPT Plus and the OpenAI API are separate products on separate billing. Plus only covers the ChatGPT web and app interface, it carries no API key and no API credits, so there is no way to point the AI Agent node at it, and you are not missing an option in n8n.
See this:

@FelixL

the ChatGPT Plus Subscription only includes ChatGPT with the GUI, the api can only be accessed via the api key which gets priced by usage. You can only use the API with N8N.

To cut the spend, keep it on the API but swap the model. The AI Agent node’s Chat Model slot takes any OpenAI-compatible endpoint, so replace the OpenAI Chat Model sub-node with a cheaper provider like OpenRouter, Groq, or DeepSeek, or keep the OpenAI credential and set its Base URL to the provider, for example https://openrouter.ai/api/v1. For zero per-token cost, run a local model through the Ollama Chat Model sub-node.
Provider and cost comparison for n8n here:

AI Inference Providers for n8n: Pricing and Comparison

[ahmeqas]

has anyone tried thebuildguild/cli-bridge - Docker Image? According to its docs, it supports Raspberry Pi too. Will try it on this weekend

[ahmeqas]

This weekend, i had deployed cli-bridge on my home lab. I had to login using my gmail account to activate the cli-bridge for free. I got confused by the instructions to setup the .env file. But then got blog post and it worked like a charm.

Now testing it for last four days with my N8N based Whatsapp Ai Chat bot. No error / request failed so far. I have 87 requests so far which consumed 575.2k / 11k tokens with my claude pro subscription.

I tried text completion and images upload which worked like a charm.

Cli-Bridge also support OFFLINE voice transcription which I have not yet tested. It claims to use “Open Source” Whisper for voice transcription. This is the work of next weekend. Get back with updates.