Comparison
Oscilar vs ArboRule: AI agents, or logic you can read.
Oscilar's pitch is agents that handle the work. Ours is a decision a person can read. Both are legitimate; they lead to very different products.
The short answer
Choose Oscilar when
- Alert triage is your cost centre. The agent hub automates the paperwork layer — AML first-line review, sanctions and PEP hit disposition, SAR narrative drafting, regulatory filing generation — which is where a lean compliance team actually loses its hours.
- You need fraud, credit and AML on one platform. Covering all three with one vendor and one data layer is a real consolidation, not a checkbox.
- You want models fine-tuned on your data rather than a consortium average, and you would rather the vendor trained them than hire for it.
- Natural-language authoring appeals. Describing a workflow in words and having it built, validated and stress-tested with generated edge cases is a genuinely different way to work.
- A marketplace of around 100 data partners across identity, KYB, screening and permissioned data would take you a year to assemble yourself.
Choose ArboRule when
- You want to read the decision, not the agent's reasoning about it. A flow is a graph of explicit nodes; what it did is what it says it did.
- You need to evaluate before a sales call. Our price is published and our API is documented publicly; Oscilar's docs site is password-protected and there is no pricing page.
- You need to know where it runs. Oscilar publishes certifications but not its deployment model — SaaS, VPC or on-premises is not stated anywhere on the site.
- Retention has to be a number you can quote. Ours is five years by default, configurable per workspace; theirs is not published.
- The team is small and the volume is modest. Priced by decision from US$ 500 a month, self-serve, with no contract to negotiate first.
Side by side
Every row links to where it came from.
Claims about Oscilar are taken from Oscilar’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 | Oscilar | ArboRule |
|---|---|---|
| The pitchoscilar.com read 2026-07-30 | different model An agentic risk platform: AI agents that handle detection, decisions and resolution across fraud, credit, onboarding and AML, sold to banks, fintechs, credit unions and digital-asset platforms. | A decision platform. Agents and model calls are nodes you can place inside a flow, not the organising idea of the product. |
| Pricingoscilar.com/demo read 2026-07-30 | not in the product Not published. There is no pricing page, no tiers anywhere on the site, and no free trial or self-serve signup — the only route in is a demo request. | 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. |
| Documentationoscilar.com — platform read 2026-07-30 | not in the product Password-protected. docs.oscilar.com exists but is not publicly readable, so no developer or API reference can be checked before a sales conversation. | Public, including the decision and history endpoints, request and response shapes, API key scopes and a per-flow OpenAPI 3.1 document. |
| Where it runsoscilar.com/security read 2026-07-30 | not in the product Not published. The security page documents SOC 2 Type 2, ISO 27001 and PCI DSS, AES-256 at rest and TLS in transit, but never states whether the product is multi-tenant SaaS, a private cloud or on-premises, nor which regions it runs in. | 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. |
| Authoringoscilar.com — platform read 2026-07-30 | in the product A drag-and-drop low-code workflow builder plus natural-language authoring — describe the workflow and rules in words and watch them come to life — explicitly aimed at reducing dependence on engineering. | 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. |
| Rollout and testingoscilar.com — credit decisioning read 2026-07-30 | in the product Backtesting, shadow mode and controlled rollout are described as a sequence for trying changes before they go fully live, with one-click backtesting of credit policies against historical data and A/B testing of model workflow versions. Explicit version numbering and rollback are not described in their own words. | 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. 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. |
| Decision historyoscilar.com — AI agents read 2026-07-30 | different model Traceability is claimed clearly — which data was used, which models contributed, what rules fired and why the outcome was reached — and agent alerts carry the input, the agent's reasoning and the final output. Retention is not published, and querying is described through a natural-language analytics agent rather than an API. | 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. |
| External dataoscilar.com/marketplace read 2026-07-30 | in the product A public marketplace of around 100 data partners across identification, KYC and KYB, digital authentication, internal data systems, specialised risk sources, permissioned data and screening. Note that the site quotes 50-plus, 80-plus and 100-plus in different places. | 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 reviewoscilar.com — case management read 2026-07-30 | in the product A full case-management product with intelligent queues routing by case type, severity and expertise, AI ranking by urgency, bulk operations, checklists, comments and requests for information, plus AI case summaries and auto-drafted filing narratives. | 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. |
| Modelsoscilar.com read 2026-07-30 | in the product Both theirs and yours. They fine-tune foundational models on your transaction data and fraud signals rather than using generic consortium models, and separately support integrating your own custom models. | 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. AI and Agent nodes call OpenAI, Anthropic or Google models inside the decision, on your own API key. No key, no AI. |
Two different bets on where the work is
Oscilar's bet is that the expensive part of risk operations is not deciding — it is everything after. Reviewing alerts, dispositioning sanctions hits, writing the narrative for a filing, producing an explanation a regulator will accept. Their agent hub is aimed squarely at that, and if you have watched a compliance team spend its week on first-line review you will recognise the problem as real.
Ours is that the expensive part is the policy drifting away from anyone's understanding of it. That a decision must be legible to the person accountable for the loss rate, reproducible years later, and changeable without a deploy. Everything in the product follows from that: explicit node types rather than emergent behaviour, a stored trace rather than a summary, weighted versions rather than a single live model.
These are not opposites and both products contain a bit of the other. But they set different defaults, and the default is what you live with.
An agent that explains its reasoning and a flow that records its path are answering the same question with different levels of guarantee.
What could not be verified
Oscilar's documentation site is password-protected. That is a legitimate choice and plenty of good companies make it, but it puts a hard limit on what a comparison page can honestly assert.
Three things in particular are not published anywhere on their site: the deployment model — whether it is multi-tenant SaaS, a private cloud or on-premises, and in which regions; how long decision records are retained; and whether decision history is reachable through an API rather than through their analytics agent. Each of those is a question a security review will ask, and none of them can be answered before a call.
One more thing worth flagging because it affects how you read their numbers: the integration count is quoted inconsistently across their own pages — 50-plus, 80-plus and 100-plus all appear. The marketplace page's 100-plus data partners is the figure we used, since it is the one attached to the actual catalogue.
The rollout claim, read carefully
Oscilar describes backtesting, shadow mode and controlled rollout as a sequence, which is the right sequence and more than several competitors offer.
Worth noting on our side of the table: ArboRule has weighted traffic and it has replay against recorded decisions, but it does not have shadow mode. A challenger version here serves a real slice of traffic or it serves none — there is no mode where it runs in parallel and its output is discarded. If mirrored evaluation is specifically what you need, Oscilar claims it and we do not.
What we do have that is unusual is what the replay pins. A dataset built from decision history carries the response each provider actually returned at the time, so rerunning last quarter's applications against a new policy is deterministic rather than dependent on what the bureau says today. That is the difference between testing a policy change and testing a policy change plus six months of vendor drift.
Questions
How much does Oscilar cost?
Not published. There is no pricing page on oscilar.com and no plan tiers appear anywhere on the site; the only conversion path is a demo request, with no free trial or self-serve signup.
Does Oscilar run in my own cloud?
Not published either way. Their security page documents SOC 2 Type 2, ISO 27001 and PCI DSS along with encryption at rest and in transit, but never states whether the platform is multi-tenant SaaS, a dedicated private cloud or on-premises, nor which regions it operates in. If deployment model matters to your review, ask early.
What is the difference between an AI risk agent and a decision flow?
An agent is given a goal and some tools and decides how to reach it; a flow is a graph you drew, executed in the order you drew it. The agent handles variation you did not anticipate, at the cost of a path you cannot fully predict in advance. The flow does exactly what it shows, and anything it did not anticipate falls through to a rule or a person. For a decision that has to be defended, that predictability is usually the point — which is why ArboRule keeps agents as a node inside a flow rather than as the flow.
Last verified July 30, 2026 against Oscilar’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.