Skip to content
SPECIFICTION
Get listed

Ojuri

ojuri.io

Free plan Open source WebWindowsLinuxAPISelf-hosted
Visit ojuri.io
Spec score 8.4 #2 of 19 in Financial Fraud Detection Software

What it is

Ojuri is a self-hosted fraud detection engine for fintechs, payment gateways, and mobile-money platforms. It scores transactions and returns accept, review, or decline decisions, with applications able to call its POST /v1/predict API. The Real-Time Detection Agent scores payments on the authorization path using an XGBoost model served through ONNX Runtime. A Pattern Analysis Agent follows account, device, and payee relationships over time to identify fraud rings and mule networks. Decisions record reason codes and the data considered by the model in an audit log. Sentinel provides a review queue, decision stream, model registry, and analyst overrides that can become training data. Other agents monitor drift, retrain on verified labels, and check evaluations before model promotion. An optional investigation agent creates reports for declined transactions using a self-hosted language model, separate from live scoring. Ojuri says its services and transaction data remain within the operator’s infrastructure. The project is MIT licensed, with no per-call fees or rate limits.

Spec sheet

What it does
Ojuri scores payment transactions before authorization and returns accept, review, or decline decisions with reasons.Read 15 Oct 2026 · ojuri.io
Purpose
Ojuri is a self-hosted fraud detection engine that scores transactions and returns accept, review, or decline decisions.Read 8 Oct 2026 · ojuri.io
Deployment
The quickstart uses Docker Compose and requires Docker and Node 20; the investigation model is optional and downloads about 7.6 GB of weights when enabled.Read 8 Oct 2026 · ojuri.io
Integrations
The stack uses Kafka, Redis, PostgreSQL, ONNX Runtime, Docker Compose, NGINX, and Prometheus/Grafana components.Read 15 Oct 2026 · github.com
Compliance
Ojuri says self-hosting can help with data residency requirements including GDPR, NDPR, and CBN, and explicitly says it is not a compliance certification.Read 8 Oct 2026 · ojuri.io
License
The repository is released under the MIT license.Read 15 Oct 2026 · github.com
Maker
Abodunrin Ayodeji says he built Ojuri as a solo project arising from research into Nigerian mobile-money fraud.Read 15 Oct 2026 · linkedin.com
Analyst dashboard
Sentinel includes a review queue, decision stream, model registry, and analyst overrides that can become training data.Read 8 Oct 2026 · ojuri.io
Analyst workflow
Sentinel provides a review queue, live decision stream, rule editor, model registry, audit log, investigations, and user administration.Read 15 Oct 2026 · ojuri.io
Architecture
The system uses four agents: real-time detection, pattern analysis, model learning, and fraud investigation.Read 15 Oct 2026 · ojuri.io
Audit trail
Ojuri records each decision and the feature snapshot used at inference before returning the response.Read 15 Oct 2026 · ojuri.io
Benchmark result
The repository reports 34% fraud detection before retraining and 98.8% after one label-driven retraining round, with 1.1% false positives in its simulation.Read 15 Oct 2026 · github.com
Data handling
Ojuri says its services run in the operator’s infrastructure and transaction data does not leave that boundary for scoring, training, or investigation.Read 8 Oct 2026 · ojuri.io
Data residency
The maker states that scoring, training, and investigations run inside the customer’s infrastructure without transaction-data egress.Read 15 Oct 2026 · ojuri.io
Explainability
Responses include reason codes, feature contributions, observed values, model version, threshold, and audit ID.Read 15 Oct 2026 · ojuri.io
Extensibility
Custom pre-model and post-model rules can be edited without restarting and normally update within 30 seconds.Read 15 Oct 2026 · github.com
Feature extensibility
The default model exposes 64 features and supports defining additional signals without code changes.Read 15 Oct 2026 · github.com
FIA resource limit
The optional investigation service downloads about 7.6 GB of model weights and needs about 16 GB of free RAM.Read 15 Oct 2026 · github.com
Fraud rings
The Pattern Analysis Agent tracks account, device, and payee relationships over time to identify fraud rings and mule networks.Read 8 Oct 2026 · ojuri.io
Fraud-ring detection
Pattern analysis identifies shared-device clusters, relationship groups, and emerging transaction rings.Read 15 Oct 2026 · ojuri.io
Investigations
The Fraud Investigation Agent creates reports for declined transactions using a self-hosted language model and runs on a separate path from live scoring.Read 8 Oct 2026 · ojuri.io
Latest release
The homepage changelog identifies v1.7.0, dated September 28, 2026, as the current release shown there.Read 15 Oct 2026 · ojuri.io
Launch date
The project’s comparison page identifies v1.0 as released on June 7, 2026.Read 15 Oct 2026 · ojuri.io
Learning loop
Reviewer overrides, chargebacks, and disputes can become labels for later retraining.Read 15 Oct 2026 · github.com
Limit: hosted offering
The project does not provide a hosted SaaS deployment or per-transaction service; customers operate the infrastructure themselves.Read 15 Oct 2026 · ojuri.io
Limit: maturity
The builder describes Ojuri as version 1 and solo-built, with its shipped model trained on public and synthetic data rather than customer traffic.Read 15 Oct 2026 · linkedin.com
Limit: operations
PAA must run as exactly one instance because it keeps the payment network graph in memory.Read 15 Oct 2026 · github.com
Limit: roadmap
The roadmap still lists Helm, Terraform, Python client support, canary rollouts, OAuth 2.0, mTLS, and Stripe, Adyen, and Plaid connectors as planned work.Read 15 Oct 2026 · github.com
LLM investigations
FIA uses a self-hosted Phi-3 model to create structured reports for declined transactions on a separate Kafka path.Read 15 Oct 2026 · ojuri.io
Model learning
The Model Learning Agent monitors model drift, retrains on verified labels, and gates model promotion on evaluation checks.Read 8 Oct 2026 · ojuri.io
Model promotion
Candidate models pass shadow and statistical gates before promotion to active status.Read 15 Oct 2026 · ojuri.io
Node SDK
The published @ojuri/sdk package provides typed prediction calls, retries, and webhook-signature verification for Node callers.Read 15 Oct 2026 · github.com
Pattern analysis
PAA maintains one-hour and 24-hour velocity windows and graph-based receiver metrics asynchronously.Read 15 Oct 2026 · ojuri.io
Performance
The reference benchmark reports approximately 6 ms server-side p99 latency, but the maker says this is not an SLA.Read 15 Oct 2026 · ojuri.io
Praise: architecture
A LinkedIn commenter praised the drift-aware retraining design and its statistical promotion gate.Read 15 Oct 2026 · linkedin.com
Public API
The primary integration is POST /v1/predict with six required transaction fields and roughly 40 optional context fields.Read 15 Oct 2026 · ojuri.io
Python requirement
Running the learning or investigation services outside Docker requires Python 3.11 exactly.Read 15 Oct 2026 · github.com
Real-time detection
The RDA service combines Fastify, Zod validation, Redis features, ONNX inference, rules, auditing, and Kafka publishing.Read 15 Oct 2026 · ojuri.io
Real-time scoring
The Real-Time Detection Agent scores payments on the authorization path using an XGBoost model served through ONNX Runtime.Read 8 Oct 2026 · ojuri.io
Security controls
The documentation describes API keys, dashboard accounts and roles, signed webhooks, and idempotency keys for safe retries.Read 8 Oct 2026 · github.com
Self-hosting
Ojuri runs on the operator’s servers with Docker Compose and does not require a SaaS account.Read 15 Oct 2026 · github.com
System requirements
The quickstart requires Docker 20.10 or newer, Compose 2.24 or newer, and Node 20 or newer for dashboard or host development.Read 15 Oct 2026 · github.com
Target users
Ojuri describes itself as fraud detection for fintech and discusses use by fintechs, payment gateways, and mobile-money platforms.Read 8 Oct 2026 · ojuri.io
Webhooks
Ojuri can send signed decision webhooks with retry behavior.Read 15 Oct 2026 · github.com
Windows
The project provides separate Windows setup instructions for running Ojuri.Read 8 Oct 2026 · github.com

Plans and prices

Free Free No contract or per-transaction fees Self-hosted deployment · no rate limits · customer-operated infrastructure Read 15 Oct 2026 · ojuri.io
Ojuri Free MIT licensed · no per-call fees · no rate limits Read 8 Oct 2026 · ojuri.io

Every Ojuri plan and price →

Strong points and limits

Strong points

  • Returns accept, review, or decline decisions
  • Tracks account, device, and payee relationships
  • Audit log records reason codes and model input
  • No per-call fees or rate limits
  • Offers a Node SDK with typed requests

Limits

  • Self-hosting requires Docker and Node 20
  • Investigation model downloads about 7.6 GB of weights
  • Investigation model is optional

Who it suits

Ojuri is aimed at fintechs, payment gateways, and mobile-money platforms that need transaction scoring within their own infrastructure. It also suits teams that want an audit trail and analyst review tools.

Verdict

Ojuri combines transaction scoring, relationship analysis, and analyst workflows in a self-hosted system. Its optional investigation model has a substantial stated download, so teams should account for that setup if they enable it.

Where Specifiction places it

Alternatives to OjuriAll alternatives →