Amigo AI — The first artificial intelligence for the Commodore Amiga

[AmigoKI]

Hello everyone!

I'd like to introduce a project I've been working on for over a year: Amigo AI — an AI assistant built specifically for the Amiga community.

What is Amigo?

Amigo is an artificial intelligence that truly knows the Commodore Amiga. Not superficially like ChatGPT, but with real deep knowledge — trained on thousands of Amiga documents, hardware references, programming tutorials and demo scene guides.

What can Amigo do?

Amiga Expert Knowledge — Questions about Custom Chips (Agnus, Denise, Paula), AmigaOS, Workbench, hardware expansions and the entire Amiga history
Code Assistance — Generates and explains code in 68000 Assembly, AMOS Basic, Blitz Basic, C and ARexx
Integrated Code Editor — Syntax highlighting for all Amiga languages, split view for chatting and coding simultaneously
AROS Emulator — Test your code directly in the integrated Amiga emulator, no manual copying needed
Knowledge Base — RAG system with thousands of Amiga documents for precise answers
Three Themes — Dark, Light and of course the authentic Amiga Workbench Retro theme

Who is Amigo for?

Anyone interested in the Amiga — whether you're an active developer, hardware tinkerer, demo coder or nostalgic user dusting off your A500 from the attic.

Example questions you can ask:
- "Write me a Copper list for a rainbow effect"
- "How does sprite collision work in Blitz Basic?"
- "What expansions do you recommend for an Amiga 1200?"
- "Create a starfield demo in AMOS"
- "What exactly does the Blitter do and how do I program it?"

Available as:
- Browser version
- Desktop client (Windows/Linux)
- Android app (coming soon)

Pricing:

The development, operation of the AI servers and the continuous expansion of the knowledge base unfortunately come with significant costs. To keep Amigo running and evolving long-term, I depend on your support.

- BASIC (15 EUR/month) — Full data access, AI chat, downloads
- PRO (25 EUR/month) — Everything in BASIC + code editor, AROS emulator, desktop client, Android app

Every contribution directly helps with further development — new features, better answers, more Amiga knowledge in the database.

Launch: April 5, 2026 at https://retroprojects.de

I look forward to your feedback, questions and of course criticism. That's the only way Amigo gets better!

Best regards,
Heiko / AmigoKI
https://retroprojects.de

[AmigoKI]

@NinjaCyborg: Good question! Tools like Cursor and Claude Code are great general-purpose coding assistants, but they lack deep Amiga-specific knowledge. Ask any generic AI about Copper list timing, Blitter channel priorities, or how to set up an A1200 interrupt handler — you'll get vague or wrong answers.

Amigo AI is built on a custom RAG system (Retrieval-Augmented Generation) with over 49 million knowledge chunks from Amiga-specific sources — hardware references, programming guides, magazine archives, demo scene tutorials. It knows the difference between OCS, ECS and AGA, understands the Custom Chips at register level, and can help with 68000 Assembly, AMOS, Blitz Basic, C and ARexx with actual Amiga context. The base LLM is an open-source model (Qwen3 8B) specifically configured for this domain — not a generic ChatGPT with a prompt in front of it.

Plus it has a built-in AROS emulator and code editor — you write code, test it, and get AI help all in one place. That's something Cursor can't do for Amiga development.

@walkero: Thanks for the interest! The knowledge base currently focuses on classic AmigaOS (1.x-3.x), hardware documentation and 68000 development. AmigaOS 4, MorphOS and AROS are covered to some extent through community documentation in the RAG, but the depth isn't as strong yet as for classic Amiga topics. Expanding coverage for these platforms is definitely on the roadmap — if you have good documentation sources for OS4/MorphOS, I'd be happy to include them.

I'm also currently developing a Dual-LLM system with a specialized coding model for programming questions, which will significantly improve code generation quality. Stay tuned!