insurance-claims-triage-mcp

AI-powered insurance claims triage MCP server. Analyze claims for fraud risk, determine priority and routing, estimate reserves, and process FNOL (First Notice of Loss) intake.

Tools

| Tool | Description | |------|-------------| | triage_claim | Full triage analysis — fraud scoring, priority, complexity, reserve estimate, and routing in one call | | detect_fraud_signals | Deep fraud analysis with specific indicator identification and investigation recommendations | | route_claim | Determine optimal adjuster assignment based on claim characteristics and workload | | estimate_reserve | Calculate initial reserve using industry benchmarks with jurisdiction and representation adjustments | | generate_fnol_summary | Parse unstructured loss descriptions into structured FNOL with coverage triggers and action items |

Installation

npx insurance-claims-triage-mcp

Configuration (Claude Desktop / Cursor / VS Code)

{
  "mcpServers": {
    "insurance-claims-triage": {
      "command": "npx",
      "args": ["-y", "insurance-claims-triage-mcp"]
    }
  }
}

Example Usage

Triage a new claim

Triage this auto collision claim: $45,000 claimed, reported 3 days after loss, 
policy is 2 months old, 1 injured party, no prior claims.

Fraud detection

Check fraud signals: claimant has 4 prior claims in 2 years, policy is only 15 days old, 
witnesses are family members, claim amount is $120,000.

Reserve estimation

Estimate reserve for a bodily injury claim in Florida, $80,000 claimed, 
attorney represented, 2 injured parties, fraud score 25.

FNOL processing

Process this FNOL: "Called in at 3pm - policyholder says their basement flooded 
overnight on March 15. Water heater burst. Carpet, drywall, and furniture damaged. 
Estimate around $30,000. No injuries. Cannot use basement. Policy HO-2847193."

Supported Claim Types

• auto-collision — Vehicle collisions
• auto-comprehensive — Theft, vandalism, weather damage to vehicles
• property-fire — Fire and smoke damage
• property-water — Water and flood damage
• property-theft — Burglary and theft
• property-wind — Wind, hail, and storm damage
• bodily-injury — Third-party injury claims
• workers-comp — Workplace injury claims
• general-liability — General liability claims
• professional-liability — E&O and professional liability
• product-liability — Product defect claims
• medical-malpractice — Medical malpractice claims

Fraud Detection

Analyzes 15 evidence-based fraud indicators across 5 categories:

• Behavioral — Staged accidents, social media contradictions
• Financial — Financial stress, inflated damages, policy stacking
• Temporal — Suspicious timing, reporting delays
• Documentary — Inconsistencies, over-preparation
• Network — Related witnesses, provider patterns

Powered By

ClaimDesk AI — Full claims adjudication platform with real-time triage, team routing, settlement analysis, and document management. Starting at $500/mo for claims operations teams.