Quickstart

1. Write a prompt

Open a new board and focus the AI bar at the bottom. Describe the system you want — the more you say about stack, scale, and constraints, the more accurate the result. A good first prompt:

Notice the prompt names the scale (“5,000 concurrent users”). Skeema’s prompt enricher reads that and automatically adds the infrastructure a system at that scale needs — a load balancer, a cache, read replicas — so you don’t have to spell out every box.

2. Generate

Press ⌘/Ctrl + Enter or click Generate. Skeema parses the prompt, infers the services, detects which connections are asynchronous, and lays the diagram out automatically. In a few seconds you get a connected architecture with domain zones (Clients, Gateway, Core Services, Data & Events) and labeled edges.

3. Refine

• •Drag nodes to reposition; connect handles to add edges; double-click a label to rename.
• •Select a node to open its side editor (technology, group, properties).
• •Ask the AI to change the diagram in place — e.g. “add a CDN in front of the web app” — and it smart-merges without destroying your layout.
• •Press ⌘/Ctrl + Shift + L to re-run the auto-layout at any time.

4. Analyze

1. 1Run a load simulation — dial the user count up and read per-path P50/P95/P99 latency, the A–F score, and the named bottleneck.
2. 2Open the cost estimator to map real cloud SKUs to each node and see a monthly bill.
3. 3If you have database nodes, derive an ER schema and run the 9-rule schema-health check.

5. Derive & export

Hover any node and use the action strip to derive views: an ER diagram from a database, a sequence flow for a service, or production code. Every derived board remembers its source. When you’re ready to hand off, export PNG, SVG, PDF, JSON, or a generated docs.md.