Build
Assemble sources, transforms, routers, and destinations. Save compiles the graph. The first run does not wait on a separate publish ritual.
Platform
The integration engine, the workbench, and the control plane.
Sirius is organized around a simple idea: integrations should be understandable under pressure. What arrived, what changed, where it went, and whether you can send it again.
Studio
A polyglot workbench: samples, golden tests, milestones, collaboration, a project terminal, and SQL against a live connector.
Connect
Named, reusable, tenant-scoped endpoints. Probe dry-run or live before a flow depends on them.
Monitor
Message search, structured logs, a dead-letter queue, and replay that writes new history instead of rewriting the past.
Govern
RBAC, SSO, API tokens, audit, and secrets that stay referenced — in the UI, the API, and the agent at once.
Deploy
One static binary. Compose for day one. Kubernetes for production. Upgrade and roll back the app and the data layer together.
Native AI agent
A first-party agent that uses the live control plane — not a separate copilot product. Ask it to draft a flow, bind connectors, or replay a failure; it does the work in minutes under your permissions.
Architecture in one sentence
The broker is not the audit log. Postgres is the product record. NATS JetStream (or RabbitMQ) moves work; Postgres proves what happened.
Read the concepts