Comparison
Nected vs ArboRule: general automation, or risk decisions.
Nected and ArboRule both publish a price and both let a non-engineer edit a rule. They part company on what the decision is expected to be about.
The short answer
Choose Nected when
- Your rules are not only about risk. Nected's own use-case list runs to dynamic pricing, discounts, payouts, taxation, payment routing, lead allocation and loyalty — risk is one entry among two dozen.
- You want rules in Git. Rules, workflows, datasets and variables sync to a real GitHub repository with branches, pull requests and merge protection, so policy changes go through the pipeline you already have.
- You need to self-host on a small budget. Free and self-hosted tiers are documented with Docker Compose and Kubernetes via Helm, alongside Terraform, EKS and AWS Marketplace paths.
- Low volume matters more than depth. The published ladder starts at a free tier with 10,000 API calls a month and rises through roughly US$ 399 and US$ 849, metered on invocations with per-editor seats.
- You want JavaScript, not Python. Conditions, results and workflow nodes take JavaScript, with an Excel-like formula editor beside it.
Choose ArboRule when
- A decision has to wait for a person. Nected's approval flow gates publishing, not decisions; a running rule can only pause on an external webhook signal, with no reviewer queue behind it.
- You need to test against what actually happened. Bulk testing in Nected runs against test data you supply; ArboRule builds datasets out of real past decisions with every provider response pinned.
- You want to move traffic gradually. Weighted rollout across published versions, sticky per case, rather than promoting a version to an environment all at once.
- The decision runs a model. ONNX models execute inside the flow, immutable and hash-verified; Nected documents no custom model execution.
- The vocabulary matters. Cases, referrals, SLA clocks, reviewer queues and decision history exist because the product is about credit, fraud and onboarding, not because they were configured that way.
Side by side
Every row links to where it came from.
Claims about Nected are taken from Nected’s own documentation, pricing and repositories — never from a review site, and never from us. Follow the link in the left column to check any of them.
| Criterion | Nected | ArboRule |
|---|---|---|
| What it is fordocs.nected.ai — use cases read 2026-07-30 | different model General-purpose business automation. The published use-case list spans dynamic pricing, discounts, payouts, taxation, payment routing, lead allocation, loyalty and personalisation alongside scoring, underwriting and fraud. | Credit, fraud and onboarding decisions specifically. The primitives — referral, case, reviewer queue, decision history — are risk vocabulary, not configuration. |
| Pricingnected.ai/pricing read 2026-07-30 | in the product Published and self-serve: free at 10,000 API calls a month, then roughly US$ 399 and US$ 849 tiers, metered on invocations with overage and per-editor seats. Private cloud and on-premises are separate tiers. | US$ 500 a month on the Startup plan, including 50,000 decisions across Sandbox and Live. Priced by decision, not by seat. Building and testing is free up to 10,000 decisions a period. |
| Where it runsdocs.nected.ai — self-hosting read 2026-07-30 | in the product Cloud, private cloud and on-premises. Self-hosting is documented in depth — Docker Compose for evaluation, Kubernetes with Helm for production, plus Terraform, EKS and AWS Marketplace paths. | Hosted only. There is no self-hosted, on-premises or bring-your-own-cloud option, and no published certification — the controls are an append-only audit log, role-based access, scoped API keys, configurable retention and SSO/SCIM. |
| Authoringdocs.nected.ai — rule types read 2026-07-30 | in the product Five surfaces: simple rule, decision table, ruleset, rule chain and decision tree, with an Excel-like formula editor and a JavaScript editor for conditions, results and workflow code nodes. | A visual canvas with 22 node types. Rule, Decision Table, Scorecard and 2D Matrix are grid editors a policy owner can work in; operand fields also accept Python, and the Code node is Python only. |
| Versioningdocs.nected.ai — version control read 2026-07-30 | in the product Draft, in review, published and archived, with an explicit rollback that replaces the current draft with an earlier version, staging-before-production promotion, and optional Git sync with branch protection. | A version is DRAFT, PUBLISHED or ARCHIVED. Published versions are immutable — you duplicate to a new draft — and publishing can require an approved review. There is no one-click rollback: you move traffic back. |
| Staged rolloutdocs.nected.ai — environments read 2026-07-30 | not in the product Not published. Promotion runs staging to production as a whole; no percentage rollout, canary or champion-challenger appears in the documentation. | Integer traffic weights across published versions, summing to 100, sticky per grouping_id so an application keeps the version it started on. No shadow mode: a challenger serves real traffic or none. |
| Decision historydocs.nected.ai — audit log API read 2026-07-30 | in the product A real audit trail, not a preview. With input and output auditing enabled, the search API returns input, output, execution id, execution time and trace id, filterable by type, status, environment, user and date. Retention is not published. | Every run stores its input, output, the version that produced it and a node-by-node trace, including each external call's URL, status and latency and each model's SHA-256. Queryable at GET /history/api/v1/decisions. Retention is per-workspace, five years by default. |
| Backtestingdocs.nected.ai — bulk test API read 2026-07-30 | different model Bulk testing against data you supply by CSV, Sheets or database query. Replaying recorded production decisions with their original external responses is not documented. | Build a dataset from decision history — past requests as inputs, recorded outputs as expectations, every external call pinned to the response the provider actually returned — then run it against any version. Up to 500 cases over a 366-day window, executed serially. |
| External datadocs.nected.ai — REST API node read 2026-07-30 | in the product A REST node covering the usual verbs with Bearer, API key, basic and OAuth 2.0 auth, response caching and continue-on-error, plus first-class connectors for Postgres, MySQL, Oracle, SQL Server, Snowflake, MongoDB and Redshift. | 28 packaged providers, six databases, and any HTTPS/JSON API through a Custom Connection. You bring your own credentials for each: the platform orchestrates the calls, it does not resell the data. |
| Human reviewdocs.nected.ai — wait and resume read 2026-07-30 | different model The approval flow is design-time maker-checker over publishing. At runtime a workflow can pause on a wait-and-resume node until an external webhook signals it; no reviewer queue or case interface is documented. | A Manual Review node suspends the decision and puts it in a reviewer queue; Case nodes record work without blocking the run. Case queues carry SLA policies pinned onto each case at creation, with business-hours clocks and escalation rungs. |
| AI and modelsdocs.nected.ai — AI agent node read 2026-07-30 | different model An AI agent node calls OpenAI, Google AI, Anthropic or Vertex for responses, text analysis and document classification. Executing your own trained model — ONNX, scikit-learn or otherwise — is not documented. | AI and Agent nodes call OpenAI, Anthropic or Google models inside the decision, on your own API key. No key, no AI. ONNX models up to 32 MB, immutable once uploaded and re-verified by SHA-256 at scoring time, executed in-process. One graph input, CPU only. |
Two products that look alike from a distance
Both have a visual editor, decision tables, versioned rules, an API, published pricing and a free way to start. Read the feature lists side by side and you would struggle to separate them. The difference is not in the list; it is in what the list is a list of.
Nected is a rule engine for a company's rules — any of them. Its documentation puts discounting, payout logic, tax calculation, payment routing and loyalty tiers on the same page as underwriting and fraud, and that is a deliberate and defensible position. A single place for every conditional in the business is a real thing to want.
ArboRule is narrower on purpose. It assumes the thing being decided is an application, a customer or a transaction; that some of them will be referred rather than answered; that someone will eventually ask why one of them went the way it did. Those assumptions are what produce a reviewer queue, an SLA clock and five years of queryable decision history — and they are wasted on a discount rule.
A general engine can express a credit policy. It just will not have opinions about what happens to the applications it refers.
The referral gap
The clearest place the two diverge is what happens when the answer is neither approve nor decline.
In Nected, publishing a rule can go through maker-checker approval, which is genuinely useful governance. But that gates the change, not the decision. A running workflow can pause at a wait-and-resume node until something external calls back with a resume id — which means the reviewer interface, the queue, the assignment, the ageing and the escalation are all yours to build.
In ArboRule a Manual Review node suspends the run and creates a review case that lands in a queue, and a Case node records work alongside a decision without blocking it. Case queues carry SLA policies with business-hours clocks and escalation rungs, and the policy is pinned onto each case when it is created, so tightening the target next quarter does not retroactively rewrite a promise already made.
If most of your decisions are automatic and referrals are rare, that machinery is overhead. If referrals are ten or twenty percent of volume — the normal range in credit and onboarding — it is most of the operational work.
Where Nected is straightforwardly better
Git. Nected's rules, workflows, datasets and variables sync to a real GitHub repository, with branches mapped to environments and merges governed by the same protection rules as your code. If your organisation's change control lives in pull requests, that is a better answer than any product-specific publishing workflow, ours included.
Self-hosting, too. Docker Compose, Helm, Terraform and Marketplace installs are documented properly, at a price point most on-premises offerings do not reach. ArboRule is hosted and has nothing to say here.
Questions
Is Nected a credit or fraud decisioning platform?
Not specifically. Nected positions itself as a business-friendly rule engine and its own documented use cases span pricing, discounts, payouts, taxation, payment routing, lead allocation and loyalty alongside scoring, underwriting and fraud. You can certainly build a credit policy in it; what you will not find is risk-specific machinery like a referral queue, an SLA clock or replay against recorded decisions.
Does Nected publish its pricing?
Yes, which is unusual in this category and worth crediting. There is a free tier at 10,000 API calls a month, paid tiers around US$ 399 and US$ 849 with invocation allowances and overage, per-editor seats, and separate private-cloud and on-premises options. Most of the enterprise decisioning vendors on this site publish nothing at all.
Can Nected run a machine learning model inside a rule?
Not one you trained. There is an AI agent node that calls OpenAI, Google, Anthropic or Vertex for responses, text analysis and document classification, but nothing in the documentation describes uploading and executing your own model file. If a score from your own model has to be produced inside the decision, that is a gap to check before committing.
Last verified July 30, 2026 against Nected’s public documentation. Products change and pages move; if something here is out of date, tell us at hello@arborule.com and we will correct it.