Core concepts

The project model in My Architect is a single tree that everything else attaches to. Here are the core entities.

Hierarchy

A project is a tree of nodes: initiatives → epics → features → tasks (the levels depend on the preset). Every node has a status — draft, todo, in-progress, done, blocked — and a roll-up: a parent's status is derived from its children, so the top level always reflects the real state of the work.

Requirements

Nodes carry requirements of six types:

• FR — functional. What the system does, plus the acceptance checklist that proves it's done.
• NFR — non-functional. A measurable quality bar (p95, uptime) you can actually test and fail against.
• SAR — significant architecture decision. A lightweight ADR on the node: the choice, alternatives and trade-offs, so nobody re-litigates it later.
• ASR — architecturally significant requirement. A requirement that carries enough architectural weight to shape the design, kept visible so it isn't lost among ordinary FRs.
• CON — constraint. An external limit you must design inside — budget, stack, timeline, legal — flagged fixed or negotiable.
• OQ — open question. An unresolved unknown that blocks the node from going done until it's answered — so your green stops being a lie.

Requirements have a priority and a status (draft / approved / deprecated), and they show up right in the node panel.

Releases

Work can be grouped into releases — a slice of what ships in a given delivery. A release is assembled from nodes in the tree rather than kept as a separate list, so it never drifts from the real model.

Diagrams

All diagrams live on one canvas (React Flow) with auto-layout — you don't drag boxes around by hand:

• WBS — work breakdown, decomposition of the work;
• User Story Mapping — a map of user stories by release;
• Event Storming — domain events, commands and aggregates;
• Box — nested containers in the C4 spirit.

Docs

Any node can have a markdown document attached — the source of truth for that feature. It's docs, not tickets, that hold the context an AI agent later works with.

How to connect an agent — see Connecting an AI agent.