signalDEV Community2026-09-25
Building a fraud investigation agent on TigerGraph for Hacker House Goa
The article details a fraud investigation agent built on TigerGraph and LangGraph for the Hacker House Goa challenge. It mimics an analyst's workflow by querying card history, device and billing regions, and past case decisions, and uses a LightGBM classifier trained on 5,565 closed investigations. The agent found a 28-card ring via a connected components query and generates alerts beyond the 20 benchmark cases.
- for who
- Engineers building fraud detection agents or graph-based investigation workflows.
- why now
- TigerGraph's new MCP server enables agents to query graphs, a timely fraud tool.
- what changes
- They can adopt a pattern where an agent uses graph queries and past case memory to decide when not to act.
- to do
- Build a similar agent using TigerGraph MCP server with allowlisted tools and a LangGraph workflow.
key points
- 590,742 transactions, 144,432 identities, 5,565 closed cases from IEEE-CIS data
- Agent runs LangGraph workflow, queries TigerGraph via MCP server with four tools
- Connected components query exposed a 28-card fraud ring on a Samsung phone profile
#tigergraph#langgraph#fraud detection#agent workflow4 sources · confidence medium
score
score 8 out of 10. 0-10: how dense the facts are, multiplied by how much you can do with them after reading. 8+ means the topic's evidence bar is met: benchmarks and availability for a new model, amount and investors for a funding round, revenue figures for a solo-money story. Below 5 an item does not enter the digest. A press release scores 3 or less, a reprint loses 2, anything older than 14 days loses 1, a headline that misleads loses 3.
read the source