Over the past few years I have fallen in love with domain-driven design and the workshops and processes around it. One of my favourite are domain discovery techniques like EventStorming.
For those that don’t know EventStorming is a workshop for collaborative exploration of complex business domains. It’s a great way to gain a shared understanding of your systems between many different stakeholders.
When you finish your EventStorming session, you can end up with a wall of post-it notes, a shared understanding and a flow of information.
I always wondered… can we get more value from EventStorming and it’s output? What if we can take this information capture it, and reference it easier in the future (without some random Miro board etc).
So I started to experiment… can we use our friend AI to take an image of an EventStorm and transform it into a EventCatalog?
What is EventCatalog
For those that don’t know EventCatalog is a MIT open source tool that I build and maintain. It’s goal is to help people bring discoverability to their event-driven architectures. It has concepts like domains, services and messages at the core, and can be used to document various different architectures for various different stake holders.
EventCatalog can visualize your architecture and also your business processes (flows) using commands, events, services (etc).
Recently I have been working on the EventCatalog MCP server which exposes a set of tools to help developers and product folks ask questions about their architecture.
But I started to wonder… can we capture value earlier in the cycle? What if we could capture the output of an EventStorming session and transform it into a visualization, documentation and layer of governance for teams?
So I started to experiment and explore, and the results were pretty amazing!
EventStorm to EventCatalog
It seems LLMs are pretty good at understanding a given image (EventStorm) and following a set of instructions (tool) to create a documentation.
Here is a quick demo of it in action, and how to get started
In my example I used this EventStorm example, a simple model with some actors, commands and events.
Next, I give this image to Claude Code (MCP client) and told it to create my EventCatalog. A few minutes later, I had a working catalog based on the EventStorm session.
Seems the results are pretty amazing, and you could use this results as a base-line for any documentation going forward. Of course the LLM will make things up along the way, but as a starting point it could be valuable?
Anyway, if you want to try this yourself I have made a quick GitHub repo with some instructions. https://github.com/event-catalog/eventstorm-to-catalog, love to know what you think, and if you get any value from this.
