Category: Work & Productivity

Inspiration

Enterprise architects work across a sprawl of disconnected tools: Visio for diagrams, Confluence for narrative, Excel for inventories, Git for ADRs, and a CMDB for the system of record. Nothing connects, and basic questions like "what systems depend on this data entity?" or "which decisions shaped this application?" take days to answer. TOGAF's Content Metamodel already defines what an EA function should model, but there's no lightweight tool that just uses it. LevelUP EA fills that gap.

What it does

LevelUP EA is a TOGAF-inspired workspace where architects catalog information systems and their owners, model the application components, data entities, and technology components that make them up, and capture Architecture Decision Records linked directly to the systems they govern. An AI chat sits on top of the whole model, so an architect can ask "which systems handle PII?" or "summarize the decisions affecting our payments platform" and get a grounded answer. Everything is connected through the metamodel, so navigating from a system to its components to the decisions that shaped them is one click away.

How we built it

Built entirely on MeDo's PaaS. The platform provided login and user management, hosted LLM access for the AI chat, and the hosting and runtime, none of which had to be coded from scratch. The approach was to treat MeDo like a strong junior architect: describe the user, anchor on the TOGAF Content Metamodel, and iterate in vertical slices, getting one entity working end-to-end before moving to the next. AI chat came last, once there was real data for it to ground against. Challenges we ran into The hardest part was keeping cross-entity relationships coherent: bidirectional links across systems, components, data entities, and decisions that don't break when something is deleted or superseded. This is where most homegrown EA tools fall apart. The other challenge was vocabulary discipline with the LLM. Paraphrasing TOGAF concepts produced muddled output, but using the actual terminology consistently (Application Component, Data Entity, Architecture Building Block) jumped the quality significantly.

Accomplishments that we're proud of

A working EA tool, live at a real URL, built in a day instead of a quarter. The metamodel holds together, the AI chat actually answers questions against real architecture data rather than hallucinating, and zero project time went to plumbing like auth, hosting, or LLM wiring. This would normally be a multi-month custom build or a six-figure platform procurement.

What we learned

The bottleneck on EA tooling was never the model (TOGAF answered that decades ago); it was who had the time and budget to build it. MeDo removes both. The biggest single lesson on working with the platform: don't ask for software, ask for a model. Giving MeDo the TOGAF metamodel and letting the software follow produced a dramatically better result than describing features in isolation. What's next for LevelUP EA Three priorities. First, quality assurance: validation, referential integrity checks, automated tests, and audit trails so the app is trustworthy as a system of record. Second, deeper AI integrations: AI-assisted ADR drafting, automatic gap analysis, impact analysis for proposed changes, and natural-language editing of the model itself, moving from Q&A widget to genuine EA copilot. Third, an improved UI with visual relationship views, faster navigation across hundreds of entities, and a cleaner layout for read-heavy workflows.

Built With

• supabase
• typescript