Actually, it seems to be just a paid service; I don't think many regular users will like it, given what the web offers for free.

You need to choose a different model. For Gemini, click “Fetch models” and select, for example, “gemini-3.6-flash.”

Just one more thing: you'll probably get a “429 Too many requests” error. This happens when you simply generate a free key without providing billing information. It seems these APIs aren't meant to work completely for free.

The creator of AmigaGPT comments on this as follows:
https://github.com/sacredbanana/AmigaGPT/issues/40
https://github.com/sacredbanana/AmigaGPT/issues/1

I actually generated a key on OpenAI, and it only started working with the program after I added $5 to my account.

retrofaza, thanks, my mistake. Instead of pasting the code, I had created a Key file
Now I'm getting the error you see in the screenshot. What am I missing?

You must enter the API key in the settings

Chat / Chat Provider Settings

I created the key for Gemini, but it isn't being found or recognized. Do I need to name it something specific or place it in a specific location?

Tested on 2026.03-D03 - works good. Tested with GPT 5.6 Luna.

AmigaGPT for AROS x86_64
chat, image generation, text-to-speech

Ported to AROS from:
https://github.com/sacredbanana/AmigaGPT


First, you need to run an assign command in the system (possibly in user-startup)

assign AMIGAGPT: <selected directory>

You'll also need an API key from your provider. I tested it with OpenAI, but you can also use Gemini, Grok, etc.

Download link:
https://archives.arosworld.org/?function=showfile&file=network/misc/amigagpt.x86_64-aros-v11.zip