One platform. Every decision explained.
Loci monitors the entire customer journey in one place: onboarding, login, payments, and screening. Built from the ground up so every score decomposes into named signals, and every action leaves an evidence trail.
Every layer feeds the next.
Signals become decisions, decisions become cases, and analyst feedback flows back into better models, with governed human approval for production changes.
Six parts, one operating loop.
Real-time decisions with the reasoning attached.
Loci's decision engine (MADIE) evaluates every active model against each event and returns the decision, fraud score, and each signal's contribution. Policy precedence is explicit: decline actions, then critical signals, then score thresholds. Benchmarked at 21–49ms p95 over a 60-day history window.
API Reference{
"decision": "review",
"fraud_score": 47,
"latency_ms": 38,
"explanations": [
{
"rule": "LIFECYCLE_COMPOSITE_001",
"contribution": 9.65,
"reason": "Lifecycle anomaly
detected."
},
{
"rule": "High Amount Withdrawal Check",
"contribution": 6.95,
"reason": "Amount is
suspiciously high."
}
],
"policy": { "thresholds":
{ "review": 35, "decline": 75 } }
}
Drag the handle: the same decision as your analysts see it, and as your systems receive it.
Describe the risk. Get a working model.
Analysts describe what risky looks like in plain language. FLM turns the description into a weighted, versioned detection model: signals that accumulate toward a decision, not brittle if-then logic. Production-bound models can be shadow-tested against live traffic before going active.
See it liveDetect multiple outbound transfers when a
customer makes more than 10 transfers within
1 hour, because this velocity pattern is
consistent with account takeover.
Flag as HIGH risk signal.{
"name": "Outbound transfer velocity",
"weight": 4,
"operations": [
{ "type": "aggregation",
"name": "count_1h_by_entity",
"operator": ">", "value": 10 },
{ "type": "comparison",
"left": "direction",
"operator": "=", "right": "OUTBOUND" }
],
"status": "shadow_test"
}
Hover to explore
A structuring fan-out model on the canvas: three signal groups, weighted scoring, one threshold.
Controls mined from your own data.
Autographer mines historical data for the patterns your models miss and drafts evidence-scored candidate controls, each with estimated capture and alert load shown before shadow testing. A generation guard proofreads every draft against its evidence.
Explore Autographer
A candidate control traced to its evidence, with governance exits and outcome reconciliation.
Continuous trust, not one-time checks.
AccessGate fuses network, device, session, and behavioral evidence into one glass-box risk decision at sensitive steps such as signup, login, payments, and profile changes. It is designed for low-latency, edge-adjacent decisions; exact response time depends on deployment and configured integrations.
Explore Behavioral Intelligence{
"decision": "review",
"action": "challenge",
"risk_score": 82,
"signals": [
{ "name": "behavioral_biometrics",
"detail": "typing cadence off baseline" },
{ "name": "network_reputation",
"detail": "datacenter ASN, new to account" }
],
"session_token": "sct_91ab…"
}Screening that finds the indirect exposure.
Sanctions and PEP screening against 6+ global watchlists updated daily, including OFAC, UN, EU, and UK lists. Relationship detection flags customers connected to sanctioned entities, not just direct name matches, and every hit carries citations.
API Reference{
"match_type": "relationship",
"lists": ["OFAC", "EU"],
"connection": "director of sanctioned entity",
"citations": 3,
"queue": "relationship_review"
}Investigations that start with context.
Alert queues and case workflows link evidence across transactions, and when enabled, sessions, devices, and screening results. Analysts open a case with the decision explanation, the entity's history, and the related network already assembled.
See a live investigation{
"case_id": "CASE-20481",
"status": "OPEN",
"linked_evidence": {
"transactions": 7,
"sessions": 3,
"devices": 2,
"screening_hits": 1
},
"lineage": "complete",
"assigned": "fraud_ops_queue"
}Built for institutions that answer to regulators.
Production changes are designed for human approval, and the material workflow history is recorded.
Every new model runs observe-only against live traffic first. Promotion follows a staged path: observe, advise, enforce.
Models are versioned with a governed lifecycle. Who changed what, when, and why is always answerable.
Risk weights and decision thresholds are tunable per organization and per action type. A payment is treated more strictly than a login.
Decisions, signals, evidence, and approvals are persisted. Audit requests become queries, not projects.
Every request is bound to an authenticated organization. No tenant can read or influence another tenant's data or decisions.
Per-user export and delete paths span the full data footprint, scoped to the requesting institution.
Run it in your cloud. Or ours.
Unlike hosted-only platforms, Loci ships as infrastructure your institution can deploy. Start SaaS to prove value, then move the same platform into your own cloud account, VPC, or data center when residency or policy requires. The API contract doesn't change.
Loci hosts the production environment; you integrate through secured APIs. Live in days, ideal for proving value.
The platform runs inside your own cloud account or VPC. Customer and decision data stays in your environment, under your controls, in your region.
Dedicated deployment for institutions with strict residency or infrastructure mandates. Same platform, same API, your metal.
See the whole platform on your own data.
A 30-minute walkthrough: live decisions, model authoring, discovery, and investigation, end to end.