IntelliCompli MCP Server

Model Context Protocol server for AI agent integration with IntelliCompli compliance platform.

The IntelliCompli MCP Server enables AI agents (like Claude) to interact with the IntelliCompli AML/CTF compliance platform through the Model Context Protocol. It provides a complete set of tools for customer management, transaction monitoring, screening, risk assessment, case investigation, and regulatory reporting.

What is MCP?

The Model Context Protocol (MCP) is an open standard that allows AI models to safely interact with external systems and tools. MCP enables Claude and other AI agents to run real operations—create customers, screen for sanctions, generate reports—while maintaining security and auditability.

Installation

Install from the monorepo:

pnpm build --filter=mcp-server

This builds the MCP server into the dist/ directory.

Configuration

Configure the MCP server by setting environment variables and adding it to your .mcp.json file.

Environment Variables

• COMPLI_API_KEY (required) — Your IntelliCompli API key (format: sk_live_...). Get this from the IntelliCompli dashboard.
• COMPLI_API_URL (optional) — Custom IntelliCompli API endpoint. Defaults to the public API.

Claude Code Configuration

Add the MCP server to your Claude Code .mcp.json configuration:

{
  "mcpServers": {
    "intellicompli": {
      "command": "node",
      "args": ["packages/mcp-server/dist/cli.js"],
      "env": {
        "COMPLI_API_KEY": "sk_live_..."
      }
    }
  }
}

The server runs as a stdio-based MCP server, making it available to Claude as a set of tools.

Available Tools

The MCP server exposes 9 tool groups covering the full spectrum of compliance operations:

1. Customers

Manage customer records and profiles.

• list_customers — Search and list customers by email, risk level, status. Paginated results.
• get_customer — Retrieve full customer profile including verification status and risk details.
• create_customer — Onboard a new customer (individual or business).

2. Screening

Run real-time compliance screenings against international watchlists.

• screen_sanctions — Check against DFAT, UN, and other sanctions lists. Supports name, DOB, and country filtering.
• screen_pep — Politically Exposed Person (PEP) screening. Identifies current/former government officials and close associates.

3. Alerts

Monitor and respond to compliance alerts generated by the system.

• list_alerts — View compliance alerts. Filter by severity, status (new, investigating, resolved, dismissed, etc.).
• get_alert — Full alert details including triggering event, associated customer, and resolution history.
• resolve_alert — Resolve, dismiss, or escalate alerts with documented reasons.

4. Cases

Formal investigation case management for high-risk activity.

• list_cases — Browse investigations. Filter by status, priority, or type (alert review, KYC review, EDD, sanctions hits, etc.).
• get_case — Full case details: tasks, evidence, assigned investigator, timeline.
• create_case — Open a formal investigation into suspicious activity or compliance issues.
• update_case — Progress case status, add findings, or close with resolution type and notes.

5. Transactions

Transaction monitoring with automatic risk-scoring and regulatory flagging.

• list_transactions — View transactions filtered by customer, risk flags, or TTR (Threshold Transaction Report) requirement. Transactions >= $10k AUD are flagged automatically.
• get_transaction — Full transaction details: risk score breakdown, TTR status, review history.
• create_transaction — Record a new transaction. Auto-scored for risk and checked against TTR thresholds. Detects structuring patterns.

6. Risk Assessment

Trigger comprehensive risk reassessments.

• assess_customer_risk — Recalculate a customer's risk level based on 11+ factors: PEP status, sanctions matches, transaction patterns, geographic risk, industry risk. May trigger EDD if risk elevated.

7. Reports

Generate regulatory reports for AUSTRAC.

• generate_ttr — Threshold Transaction Report (TTR). Required for cash transactions >= $10k AUD.
• generate_smr — Suspicious Matter Report (SMR). File within 24 hours of forming suspicion of money laundering or terrorism financing.
• generate_ifti — International Funds Transfer Instruction (IFTI). Required for all international electronic funds transfers into/out of Australia.
• list_smr_reports — View previously submitted SMRs with filtering by status or activity type.

8. Analytics

Compliance metrics and overview dashboards.

• get_analytics — Aggregate compliance metrics: customer counts, transaction volumes, screening match rates, risk distribution, TTR/review stats. Specify date range.

9. Compliance & Audit

Deep compliance infrastructure: audit trails, enhanced due diligence, beneficial ownership.

• list_audit_logs — Complete audit trail. Every action (customer changes, screening runs, case updates, report generation, API usage) is logged. Essential for compliance reviews.
• list_edd_investigations — Enhanced Due Diligence (EDD) investigations. Triggered automatically for high-risk customers; involves deeper scrutiny of source of funds, business relationships, and transaction patterns.
• get_edd_investigation — Full EDD details: checklists, source of funds verification, findings.
• list_beneficial_owners — Beneficial Owner (UBO) lists for business customers. Australian AML/CTF law requires identifying anyone with 25%+ ownership or significant control.

Caching

The MCP server includes a TTL (Time-To-Live) cache that reduces unnecessary API calls:

• Customer data: 60 seconds
• Alerts: 30 seconds
• Cases: 30 seconds
• Transactions: 30 seconds
• Analytics: 120 seconds
• Screening results: 60 seconds
• Audit logs: 30 seconds
• Reports: 60 seconds
• EDD investigations: 30 seconds

Write operations (creating customers, screening, generating reports) automatically invalidate relevant cache entries to ensure fresh data.

Development

Build

pnpm build

Watch Mode

For development with automatic rebuilds:

pnpm dev

Clean Build

pnpm clean

Architecture

The MCP server consists of:

• server.ts — Core server setup. Creates the IntelliCompli SDK client, initializes the TTL cache, and registers all 9 tool groups.
• cli.ts — CLI entry point. Launches the stdio-based MCP server.
• cache.ts — TTL cache implementation with LRU eviction and automatic expiration sweep.
• tools/*.ts — 9 tool modules, each registering a group of related tools.
• index.ts — Package exports for programmatic use.

Example Usage

List High-Risk Customers

# Via Claude Code:
list_customers risk_level=high limit=50

Screen a New Customer for Sanctions

screen_sanctions first_name="John" last_name="Doe" customer_id="cust_123"

Create and Investigate a Case

create_case title="Suspicious Transaction Pattern" case_type="transaction_review" priority="high" customer_id="cust_456"

# Then update as investigation progresses:
update_case case_id="case_789" status="in_progress"
update_case case_id="case_789" status="resolved" resolution_type="legitimate_activity" resolution_notes="Transaction pattern explained by seasonal business..."

Generate a Suspicious Matter Report

generate_smr \
  activity_type="money_laundering" \
  description="Multiple large cash deposits followed by immediate withdrawal" \
  suspicion_formed_date="2026-04-09" \
  transaction_ids=["txn_1", "txn_2"] \
  grounds_for_suspicion="Pattern consistent with money laundering typology..." \
  action_taken="Account frozen pending investigation" \
  reporting_officer_name="Jane Smith" \
  reporting_officer_position="Compliance Officer" \
  reporting_officer_contact="+61-2-1234-5678"

API Reference

For full tool parameter documentation, call tools via Claude with --help:

list_customers --help
get_customer --help
screen_sanctions --help
# etc.

Or see src/tools/*.ts for detailed Zod schema definitions and tool descriptions.

Security & Permissions

• All API calls are authenticated via your COMPLI_API_KEY.
• Write operations (customer creation, screening, report generation) require explicit user confirmation in Claude Code.
• All actions are logged in the audit trail and visible in the IntelliCompli dashboard.
• MCP operates over stdio, isolating the server process and ensuring no unintended network access.

License

MIT