Skip to main content
DEKLARATIVE
Start free
Menu

Workflow engine

A real workflow engine. Plus the modeling notations your team already knows.

The workflow engine is the heart of Deklarative — the layer that makes a low-code platform stand up to enterprise workloads. We chose two notations because the market did: BPMN for deterministic processes (approvals, fulfillment, refunds, change management), CMMN for non-deterministic case-management work (investigations, onboarding, customer success). Both run on a Temporal-compatible runtime so the durability guarantees are the ones engineers expect.

BPMN 2.0

  • Full element palette — tasks, gateways, events, sub-processes
  • Exclusive / parallel / inclusive / event-based gateways
  • Boundary events (timer, error, message, escalation)
  • Sub-process and call-activity reuse
  • Approval gates wired to your task inbox

CMMN 1.1

  • Cases with stages, milestones, and discretionary tasks
  • Sentries (entry / exit conditions) with CEL expressions
  • 10 participant roles (owner, reviewer, approver, observer, etc.)
  • Discretionary task palette — case worker decides at runtime
  • Real-world cases: investigation, customer onboarding, incident, care plan

Resilience

  • Retries with exponential backoff and jitter
  • Idempotency keys (Memory + Redis-backed stores)
  • Circuit breakers per integration
  • Saga compensation flows
  • Decorators not nodes — resilience attaches to activities, not to graph topology

Task inbox

  • Personal + group inboxes with claim / delegate / unclaim
  • Form rendering inline with outcome routing
  • Four orthogonal state machines per task: lifecycle / delegation / assignment / SLA
  • SLA states — pending, near-breach, violated, met
  • 7 prioritization views (Eisenhower, RICE, MoSCoW, WSJF, ICE, KANO, Value-Effort)

Polyglot workers

  • TypeScript: Node, Bun, Deno workers
  • Kotlin: JVM workers via Ktor
  • Go: native workers — fastest cold-start, smallest footprint
  • Rust: native workers — best for compute-heavy / WASM use cases
  • Python: workers for ML / data-science activity invocation

Observability

  • OpenTelemetry traces across the workflow
  • Per-instance event history
  • Replay debugger
  • Prometheus metrics exposed by default
  • Visual debugger in Studio (step through a real instance)

Get started

Describe the system. Ship the system.

Open source under Apache 2.0. Cloud free for evaluation. Production deployments self-hosted or hosted, your call.

Start free See pricing Read the docs →