ALL SYSTEMS OPERATIONAL · AP-SOUTHEAST-2 / AP-SOUTHEAST-1
Nodeworks Platform · built in NZ

Automation with an SLA.

The integration platform you can put in front of an auditor. EDI, durable workflow automation, and on-prem reach — one product, one bill, one append-only audit trail.

Durable Postgres-backed runs · immutable versions · resume from failed step
run r_4c19 · invoice-sync · v14
LIVE
14:32:081/6trigger.file · INVOIC receiveddone 0.2s
14:32:092/6edi.validate · guide v7done 0.4s
14:32:093/6transform.jsonatadone 0.1s
14:32:104/6http.request · ERPfailed · timeout
14:32:314/6http.request · resumeddone 1.8s
14:32:325/6human.approval · ops teamwaiting
6/6webhook.emit · signedqueued
resumed from step 4 · completed steps never re-ran · every event audited

One platform. Three modules.

Everything runs on the same durable engine, under the same tenancy and role model, into the same audit trail. Turn on what you need.

EDI

Available now

The flagship. Trading-partner EDI for supply chains where a missed document costs real money.

X12 · EDIFACT · Foodstuffs XML
Versioned partner guides per document type
997 / CONTRL acks — sent on time, chased when missing
Shadow-mode migration off legacy VANs and providers

Automations

Early access

Durable workflows — triggers to steps — with the failure story designed before the happy path.

Webhook, schedule, file, email and EDI-event triggers
Resume from the failed step, not from the start
Immutable versions · instant rollback
Approval steps, monitors and per-workflow SLOs

Agent

Coming soon

Reach on-prem systems — databases, file shares, internal APIs — without opening a single inbound port.

Outbound-only TLS tunnel · nothing listens on your side
Local credentials never leave your network
Enumerated operations — no shell, no script execution
Every operation audited, presence visible in the console

Reliability is the product.

Most automation tools treat failure as an edge case. We built the failure path first — because integrations move money, and “re-run it and hope” is not an answer you can give an auditor.

01

A failed run resumes where it failed.

Every run is persisted step by step in Postgres. When a target times out at step four, you fix the cause and resume from step four — the three completed steps never execute again, so nothing double-posts, nothing double-ships. Dead-lettered runs are one click from resolution.

Trigger → Validate → Transform → Deliver → Approve
steps 1–3 · completed0.7s
step 4 · http.request failedtimeout
resume from step 4resume →
v13 · activepinned by 4,182 runs
v14 · shadow — live traffic, zero side effectsday 22 of 30
diff: identical 4,178 · benign 4 · breaking 0cutover ready
02

Test the next version against live traffic.

Shadow deploys run a candidate version alongside the active one, on real traffic, with every side effect suppressed — then semantically diff the outputs, with a noise model so timestamps and counters don't cry wolf. You cut over on evidence: a sustained window of zero breaking diffs. Blue-green, with a brain.

Durable by construction
Every run is persisted step by step in Postgres — the source of truth, not a cache. A crash or deploy mid-run resumes from the last completed step.
Immutable versions
Workflow definitions are append-only versions. A run pins the version it started on. Rollback is instant — re-point the active version. Nothing live is ever edited.
Effectively-once side effects
Per-step idempotency keys mean at-least-once delivery plus dedupe: retries never repeat a completed side effect.
Approvals in the loop
Human-approval steps inside workflows, and approval gates on activation. Every decision lands in the audit trail with who, what and when.
Monitors & SLOs
“This workflow hasn’t completed a run in X” and “error rate above Y%” as first-class alerts, with per-workflow success rates surfaced in-product.
Hard budget caps
Usage is metered on runs, never per task. Hit your cap and runs stop — the bill never surprises you.

An honest comparison.

We don't out-breadth per-task automation tools, and we don't out-configure self-hosted ones. We win on what happens after something fails — and on who is accountable when it does.

Per-task automation tools
Self-hosted workflow tools
Nodeworks Platform
When a step fails
Replay the whole run and hope
Depends on the queue and database you operate
Resume from the failed step — completed work never repeats
Pricing model
Per task — the bill grows with every step that fires
Free to license; you pay in infrastructure and ops time
Tiered on runs, unlimited steps, hard budget caps
Versioning & audit
Edits go live immediately; history is thin
Mutable workflows unless you build the discipline yourself
Immutable versions, approval gates, append-only audit
Reaching on-prem systems
Cloud APIs only
Self-host inside the network — and patch it
Outbound-only agent (coming soon) — no inbound ports, no VPN
Who carries the SLA
Best-effort
You do
We do — published SLA with service credits
EDI
Not built for it
Roll your own codecs and acks
A first-class module: X12, EDIFACT, guides, acks

Built where documents cost money.

The platform was proven in supply chain first — an industry that fires vendors over missed documents. That discipline is now underneath every workflow, EDI or not.

Trading-partner EDI
Retailers, 3PLs and carriers. X12, EDIFACT (EANCOM) and Foodstuffs XML are first-class codecs, with versioned, JSON-Schema-backed guides per partner per document type.
Acknowledgment reconciliation
997s and CONTRLs sent well inside SLA — and missing inbound acks chased automatically, ASNs within hours. Transport receipt is verified before any resend goes out.
Chargeback firewall
Every outbound document is validated against the partner’s guide before it leaves. Catch the violation on our side of the wire — before it becomes a deduction on your remittance.
Partner onboarding
Templates for the partners you actually trade with. AS2 or SFTP, test documents included, no consulting invoice at the end.

Boring where it counts.

This is mission-critical transaction data. The infrastructure underneath it is deliberately unexciting — and we only claim what we can show you.

Encryption
TLS 1.3 in transit, AES-256 at rest. AS2 payloads signed and encrypted end to end.
Audit
Append-only, always. Every document, every run, every config change — actor, before and after. Retention options to 7 years.
Isolation
Per-organisation isolation on every row, per-tenant queues and rate limits. One tenant’s bad day never becomes yours.
Availability
99.9% SLA on Business plans, 99.95% on Enterprise — backed by service credits, not apologies.
Residency
Region-pinned infrastructure in Sydney and Singapore. Your data stays in region. NZ data residency is on the roadmap.
Access & compliance
Role-based access, scoped API keys, invite-only onboarding. SOC 2 Type II program underway — we’ll claim the badge when we’ve earned it.

Switch providers with proof, not hope.

A named Nodeworks engineer owns your migration end to end. We rebuild your mappings, mirror live traffic into the platform, and semantically diff every document against your current provider's output — a noise model keeps timestamps and control numbers from crying wolf. You cut over when breaking diffs have been zero for 30 days, with us on call through go-live.

01
Mirror
We rebuild your mappings, then duplicate live traffic into Nodeworks. Zero side effects. Your partners notice nothing.
02
Diff
Every output semantically compared against your current system, every difference classified: identical, benign, or breaking.
03
Cut over
Breaking diffs at zero for 30 days. Flip the endpoint. Decommission the old contract.

Talk to us.

Onboarding is invite-only and white-glove: a named engineer, your first workflows built with you, migration owned end to end. Tell us what you're running and we'll show you the platform against your own documents.

hello@nodeworks.co.nz