Skip to main content
completed production-ai

Fraud Agent Monitor

LangGraph-based monitoring pipeline with 4 specialized fraud agents, deployed as API and dashboard services on Cloud Run.

Started: March 28, 2026
Completed: April 5, 2026

Evidence & Verification

Metrics and claims on this page are tied to the linked artifacts below (repository docs, experiment outputs, and deployment pages when available).

2
Cloud Run Services
4
Agents
4
Pipeline Stages
8/8
P0 Checks

Tags

fraud-detectionagentslanggraphfastapistreamlit

Technologies

Python LangGraph FastAPI Streamlit Docker

Overview

Fraud Agent Monitor orchestrates four task-specific agents to score suspicious financial behavior and produce analyst-ready output:

  1. Transaction Analyzer
  2. Pattern Detector
  3. Risk Scorer
  4. Report Generator

Live Deployment

API Surface

  • POST /monitor to execute the full multi-agent orchestration
  • POST /agents/{name}/invoke to test an individual agent
  • GET /health for graph and runtime checks

Validation Snapshot

Phase 3 automated checks passed 8/8 monitor scenarios, including:

  • suspicious wire execution with full trace,
  • elevated risk scoring and typology detection,
  • SAR report generation output,
  • normal payment low-risk behavior.

GitHub Status