Ojuri

Ojuri is a self-hosted fraud engine for teams that need transaction decisions and fraud operations under their own control. It is best suited to fintechs, payment gateways, and mobile-money operators with engineering capacity to run the stack. The zero-cost, MIT-licensed plan is unusually permissive, but infrastructure and model operations remain your responsibility.

Overview

Ojuri returns accept, review, or decline decisions and covers the workflow around those outcomes. It combines live payment scoring, relationship analysis, analyst review, investigations, and model management in an omnichannel system. Services run in the operator’s infrastructure, keeping transaction data inside that boundary for scoring, training, and investigation.

That control is valuable for organizations with strict data-residency requirements, including GDPR, NDPR, or CBN-related requirements. Self-hosting can support those goals, but Ojuri is not a compliance certification. Deployment also requires Docker and Node 20; enabling investigation reports adds an optional model download of about 7.6 GB.

Key features

Live decisions and fraud networks

The Real-Time Detection Agent scores payments on the authorization path with an XGBoost model served through ONNX Runtime. It fits payment flows where an accept, review, or decline result must arrive during authorization. Pattern Analysis follows account, device, and payee relationships over time, helping teams investigate fraud rings and mule networks that single-transaction rules can miss.

Auditability and analyst control

Each decision records reason codes and the data seen by the model, giving investigators a traceable audit trail rather than an unexplained score. Sentinel adds a review queue, decision stream, model registry, and analyst overrides. Overrides can become training data, connecting day-to-day review work with later model improvement.

Learning and investigations

The Model Learning Agent watches for drift, retrains on verified labels, and requires evaluation checks before promoting a model. This is a useful governance pattern for teams that can maintain labeled outcomes; it is more responsibility than a managed service that handles the lifecycle for you. The Fraud Investigation Agent creates reports for declined transactions using a self-hosted language model on a separate path from live scoring, so investigative processing does not have to hold up authorization. The 7.6 GB model download makes that option a consideration for storage and deployment planning.

Integration and controls

Applications can call POST /v1/predict. The Node SDK provides typed requests, retries, and webhook signature verification. API keys, dashboard accounts and roles, signed webhooks, and idempotency keys address common integration and operational risks. Custom rules, machine-learning models, device fingerprinting, and omnichannel support give teams room to combine policy with model signals.

Pricing

Ojuri’s Ojuri plan costs free. It is MIT licensed, has no per-call fees, and imposes no rate limits. The plan includes the product’s custom rules, machine-learning models, device fingerprinting, and omnichannel capabilities. There are no paid tiers to trade features or quotas against, so it suits teams willing to fund hosting, integration, monitoring, and maintenance themselves. Organizations seeking vendor-managed operations or commercial support should look elsewhere.

Platforms

Ojuri is offered as an API, self-hosted deployment, web dashboard, and Windows setup. The documented quickstart uses Docker Compose and requires Docker and Node 20. Windows users have separate setup instructions, while the API and Node SDK serve application integration.

Who it's for

Fintechs, payment gateways, and mobile-money platforms are the clearest fit, especially when transaction data must remain inside an operator-controlled environment. It also suits fraud teams that want analysts, model managers, and investigators working in one system. Smaller teams without Docker, Node, model-lifecycle, or operational expertise may find a hosted alternative more practical.

Pros and cons

Pros

• No usage charges or rate limits: the free MIT-licensed plan avoids per-call costs as transaction volume grows.
• Broad fraud workflow: live scoring, graph-style relationship analysis, review queues, investigations, and model promotion cover more than an API-only detector.
• Traceable decisions: reason codes and recorded input data support audit and analyst review.
• Data boundary: scoring, training, and investigation stay in the operator’s infrastructure.

Cons

• Operational overhead: Docker and Node 20 are required, and self-hosting shifts reliability and maintenance work to your team.
• Investigation storage cost: the optional language model downloads about 7.6 GB of weights.
• Not a certification: self-hosting may help residency requirements but does not itself establish GDPR, NDPR, or CBN compliance.

Alternatives

For a broader category comparison, see Payment Fraud Detection Software.

• FraudLabs Pro is the better fit when a freemium hosted service, Android or iOS access, and a 500-query monthly free tier matter more than unlimited self-hosted use.
• IPQualityScore Fraud Prevention suits teams wanting hosted and self-hosted options with a free plan capped at 1,000 lookups per month and 35 per day, plus a $99 Startup plan.
• Jube is another free, self-hosted open-source option under AGPLv3.
• Stripe Radar is preferable for businesses already using Stripe standard payments and needing included card fraud prevention.
• Ravelin targets organizations seeking a customized paid platform.
• Tazama offers free open-source software, with hosting, implementation, and support requiring resources or a contract.
• Vesta is a paid alternative for teams evaluating a commercial platform.
• DataVisor Fraud Platform is a custom enterprise option with demo-based pricing.

Verdict

Choose Ojuri if you are a fintech or payment operator that wants unrestricted, self-hosted fraud controls, explainable decisions, and an integrated analyst-to-model workflow. Look elsewhere if you want the provider to run the infrastructure, manage model operations, or supply a packaged compliance and support offering.