Cascades
System_Live
v1.0.4-stable

Deterministic orchestration with verifiable execution

Execute workflows as structured graphs, observe every step, and verify results with immutable execution proofs.

Start buildingExplore API
Trigger LLM Node Validation

[01] / Surface

One orchestration plane — APIs, workers, proofs, and integrations

Built one boundary at a time — clarity at the seams between trigger, execute, and verify.

DAG execution

Validate graphs, schemas, and ordering before anything runs.

Proofs & audit

Hashes, drift checks, and records you can verify via API.

Workers & queue

BullMQ-backed runs with inline fallback where configured.

HTTP & OpenAPI

Contract-first APIs, SDKs, and streaming run logs.

Integrations

GitHub, Slack, Jira, Stripe, and operational connectors.

Access control

Roles, sessions, and policy gates aligned with your IdP.

Live Execution: Incident Triage + Proof

Watch a workflow pass through ingest, classification, policy checks, and proof anchoring.

Cascades
Runtime online
ALERT_INGEST0.18s
complete
LLM_SEVERITY_CLASSIFIER0.42s
executing
LOGIC_GATE--
pending
STATE_COMMIT0x00...00
queued
POST /api/workflows/run
{
  "workflow_id": "example-workflow",
  "context": {
    "source": "pagerduty"
  },
  "inputs": {}
}
GET /api/runs/{id}/stream
Inspect Runs

System guarantees

Interfaces between stages—what crosses each boundary and what each node guarantees.

DAG
DAG
Payload
Branch ctx
Trace
Trace
DEFINEvalidate · compile
In
Workflow spec
Out
Validated DAG
AGENTinvoke
In
Task payload
Out
Execution result
CONDITIONbranch
In
Runtime context
Out
Route decision
APII/O
In
Request envelope
Out
Response + metadata
TRANSFORMmap
In
Record stream
Out
Normalized record
VERIFYprove
In
Execution trace
Out
Proof hash

Execution contract

Typed handoffs · not just topology

Orchestration coordinates systems through explicit inputs, outputs, and execution semantics.
The Control Plane

Deterministic Orchestration
for the AI Era.

// Cascades is a verifiable control plane ensuring AI workflows execute as defined with cryptographic audit trails.

01

Define Contracts

Specify your workflow as a deterministic DAG with explicit schema validation.

02

Execute with Policy

Gate every node execution through policy checks, approvals, and execution state.

03

Verify & Audit

Persist immutable traces with per-task content hashing for zero-trust reviews.

Live_Runtime_Inspector
LLM_NODE_GEN
EXECUTING...~0.42s
LOGIC_GATE
PASSEDauthorized

RES: { status: "authorized", claim: "sre_lead_sig_validated" }

STATE_COMMIT
PENDING0x00...00
Trace ID: d41d8cd98f00b204NODE_UP
Technical Architecture

System capabilities

// System Specifications

  • Deterministic DAG execution engine
  • Real-time state streaming (gRPC/Websockets)
  • Cryptographic proof anchoring (SHA-256)
  • Multi-tenant worker isolation
MODULE_VERIFY
01

Verifiable Execution

Every task execution is cryptographically hashed and recorded for full auditability.

View Documentation
MODULE_GRAPH
02

Deterministic DAGs

Validate dependencies, schemas, and execution order before a workflow run begins.

View Documentation
MODULE_SCALE
03

Queue + Inline Execution

Runs execute through Redis/BullMQ workers when available, with inline fallback where configured.

View Documentation
MODULE_AI
04

Workflow graph execution

Execute DAGs with explicit dependencies, retries, and task-level state transitions.

View Documentation
MODULE_GOVERN
05

Execution Verification

Compare enqueue-time definitions to live workflows and verify proof records through API endpoints.

View Documentation

Ready to deploy
your first DAG?

Get started with our CLI tool or explore the dashboard.

Initialize Workspace
Network Placement

Where Cascades Fits

Client App
Cascades
External APIs

Cascades functions as the orchestration kernel. While external systems provide data, compute, and side effects, Cascades ensures the execution graph is respected, logged, and verified.

The Problem Space

Why Cascades Exists

Standard workflow engines are execution-only. They lack the verifiable guarantees required for agentic integrity.

The Gap

Opaque execution logs that cannot be cryptographically proven or audited in zero-trust environments.

The Solution

Deterministic DAGs coupled with SHA-256 proof generation for every node transition.

Orchestration hub

WEBHOOK
✓ verified
AGENT
CONDITION
API
TRANSFORM
VERIFY
CASCADES ENGINE

Execution lifecycle

Define workflows, execute across systems, observe state, verify outputs, and replay execution when needed.

DEFINE → EXECUTE → OBSERVE → VERIFY → REPLAY

Deterministic execution

Execution paths resolve predictably across nodes and branches.

Streaming state

Each step emits runtime state as the workflow progresses.

Verification layer

Outputs are recorded and can be replayed and validated.

Execution templatesOpen builder

Incident Triage + Proof

Ingest an alert, classify severity, request approval, remediate, and anchor proof.

6 nodesExecute

AI Content Pipeline

Research, outline, draft, fact-check, format, notify, and publish with gates.

10 nodesExecute

Change Control Review

Collect diffs, run policy checks, gather human sign-off, and record outcomes.

7 nodesExecute

Webhook Enrichment

Normalize inbound events, enrich context, branch on risk, and notify owners.

6 nodesExecute

Integrated Ecosystem — Connect via APIs & Webhooks

Cascades connects external systems through API calls, webhooks, and workflow tasks documented in the integrations reference.

All integrations execute through verifiable workflow runs with structured logs and proof records.

Read Integration Guide
12
Reference integrations
Jira
StatusAPI / Webhook
View Integration Pattern
ServiceNow
StatusAPI / Webhook
View Integration Pattern
PagerDuty
StatusAPI / Webhook
View Integration Pattern
Slack
StatusAPI / Webhook
View Integration Pattern
GitHub
StatusAPI / Webhook
View Integration Pattern
GitLab
StatusAPI / Webhook
View Integration Pattern
Stripe
StatusAPI / Webhook
View Integration Pattern
Zapier
StatusAPI / Webhook
View Integration Pattern
n8n
StatusAPI / Webhook
View Integration Pattern
Kubernetes
StatusAPI / Webhook
View Integration Pattern
Make
StatusAPI / Webhook
View Integration Pattern
Pipedream
StatusAPI / Webhook
View Integration Pattern

Execution surface

Cascades is executed through the API

Every run starts with a write—then you read the same execution back as it moves through PENDING, RUNNING, and terminal state over one HTTP surface.

Trigger

API-FIRST EXECUTION

Trigger workflows with a single request.

POST /api/workflows/run
{
  "workflow_id": "example-workflow",
  "context": {},
  "inputs": {}
}
Observe

STREAM EXECUTION

Observe state as it happens.

GET /api/runs/{id}/stream
→ PENDING→ RUNNING→ SUCCESS
Explore API
CHANGELOG

Recent system updates

// Product updates, architectural deep-dives, and release milestones.
// Current branch: origin/main

2026-03-24

Verifiable RAG Pipelines

Implementing SHA-256 anchoring for long-context retrieval...

2026-02-12

Deterministic DAG Engine v1

Release notes for the core orchestration runtime...

Read More

Precision Licensing

// Scalable orchestration tiers for every stage of deployment.

Starter
$39/mo
  • 2,500 executions/mo
  • 10 workflows
  • Community support
  • Core integrations
Get started
ProRecommended
$89/mo
  • 25,000 executions/mo
  • Unlimited workflows
  • Email support
  • Analytics
  • Policy + proof controls
Get started
Business
$219/mo
  • 100k+ executions/mo
  • Drift detection
  • Advanced access controls
  • Priority support
Get started
Enterprise
Custom
  • Dedicated infrastructure
  • SLA guarantee
  • Governance controls
  • SSO/SAML
  • Deployment support
Contact Sales