Bringing vintage documentation back to life

When we talk about preserving computer heritage, we often think about hardware, motherboards we repair, cassettes we dump and archive for posterity.

But something essential is often missing: knowledge.

How can you find, without digging for 30 minutes, the exact syntax of a 6809 assembly routine?
Or that obscure trick to compile a C program on the MO5?

That is exactly why I launched retrocomputing-ai.cloud: a RAG server (Retrieval-Augmented Generation) designed to help bring machines back to life… by making them programmable again.

My approach: a very focused development assistant

For now, I am deliberately focusing on a single playground: 👉 the Thomson MO5

BASIC is already well documented almost everywhere.
But as soon as you get into:

…the information becomes scattered, sometimes contradictory, and often difficult to track down.

And let’s be honest: I have an MO5 right here… that helps 😄

The idea is simple:

ask a question in natural language,
and let the RAG find the right page, the right excerpt, the right explanation.

Eventually, I’d like to extend the platform to other legendary machines: Commodore 64, Apple II, Amstrad CPC…

A quick word on the technical side (without diving too deep into the cables)

The server relies on a fairly classic RAG pipeline, but designed to remain autonomous and low cost.

Today:

Why this choice?

👉 almost zero cost,
👉 independence,
👉 and the ability to iterate quickly.

Eventually, I will probably move to embeddings generated by more recent AI models (better understanding, better robustness), but for now, I like this approach: no surprise bills.

Whatever engine I use, the goal stays the same: provide a useful and sourced answer, rather than a convincing hallucination.

Using the MO5 index directly in your IDE (Copilot, Augment, Claude…)

Thanks to the MCP (Model Context Protocol), your AI can plug into my knowledge base as if nothing special was happening… and suddenly become a MO5 specialist.

1️⃣ Installing the MCP server

  1. Clone the repository:
git clone https://github.com/thlg057/mo5-mcp-server.git
cd mo5-mcp-server
  1. Preparation

Make sure you have Node.js installed.

On Linux/macOS, make index.js executable:

chmod +x index.js

2️⃣ Configuring the MCP client

Add this to your configuration JSON file (Copilot, Augment, Claude Desktop, etc.):

{
  "mcpServers": {
    "mo5-rag": {
      "command": "node",
      "args": ["C:\\your\\path\\to\\mo5-mcp-server\\index.js"],
      "env": {
        "RAG_BASE_URL": "https://retrocomputing-ai.cloud"
      }
    }
  }
}

The retrocomputing-ai.com URL points to my .NET API. It handles vector search, assembles the response and acts as an interface between your IDE and the documents.

The MCP server simply acts as the bridge, cleanly, without hacks.

What’s next?

The project evolves more through experiments than through big planned milestones.

I keep improving the documentation as I go, sometimes fixing old misunderstandings, sometimes adding clearer examples. On the development side, I feel I now understand the text mode pretty well, but the graphics mode is another story: more subtle, more demanding… and I am still learning how to tame it.

So I will take the time, regularly, to update the documentation.
All my Markdown files are available with the RAG server sources, nothing is hidden, and everything can be read, improved or extended.

And if you feel like contributing, sharing documents, fixes, or ideas: get in touch. On my own, I won’t be able to keep this service alive forever, and I would really love it to become a useful tool for everyone who loves retro computing.

Enjoy 😄

Want to follow the journey?

I document this whole retro-computing adventure on my blog:
French version: https://thlg057.github.io/mo5-blog/
English version: https://thlg057.github.io/mo5-blog/en/
RAG server source code: https://github.com/thlg057/mo5-rag-server