Design systems, not drawings.
Skeema is an AI system-design workspace. Describe a backend in plain English and it generates a production-quality architecture — then derives ER schemas, sequence flows, and real code that stay linked to the source. These docs explain how it works, the concepts behind it, and how to use Skeema for real engineering tasks.
Most diagram tools draw shapes. Skeema models systems: every architecture node is a real service with a technology, traffic behavior, and failure characteristics — so the same diagram can be simulated under load, costed against real cloud SKUs, and turned into schemas and code. This documentation is written to be educational: alongside the how-to, you’ll find the underlying concepts (latency percentiles, scaling patterns, schema normalization) explained from first principles.
- ✓The architecture diagram is the source of truth; ER, sequence, and code are derived from it and stay linked.
- ✓Everything is generated from a prompt, then fully editable — drag, connect, and edit in place.
- ✓Analysis is built in: load simulation, cost estimation, and schema health — not bolted-on extras.
- ✓New to the terms? Start with Core concepts and the Glossary.
What you can build
- •System architecture — services, gateways, queues, datastores, async event flows.
- •ER schemas — entities, keys, relationships, junctions, with a 9-rule health check.
- •Sequence flows — request lifecycles with latency on every message.
- •AI-agent graphs — LLMs, tools, retrievers, memory, routers, with step playback.
- •CI/CD pipelines — build → test → deploy stages with n8n export.
Browse the docs
Who this is for
Backend and platform engineers designing or reviewing systems; product and engineering managers who need a shared picture of how a system behaves; and students or self-taught developers learning system design. No prior diagramming experience is required — if you can describe what you’re building, Skeema can draw it.