Skip to content

README.md

Latest commit

 

History

History
153 lines (113 loc) · 6.27 KB

README.md

File metadata and controls

153 lines (113 loc) · 6.27 KB

Agentic Payments

AI-powered payment processing that validates, routes, and reconciles financial transactions end-to-end using coordinated specialist agents.

Overview

The Agentic Payments application automates payment workflows by validating transactions against compliance and fraud rules, selecting optimal payment rails based on cost and speed, and reconciling payments across systems. It produces a unified decision: EXECUTE, REJECT, or ESCALATE.

Business Value

  • Reduce Payment Failures -- Pre-validation catches compliance violations and fraud signals before execution
  • Optimize Costs -- Intelligent rail selection balances speed, cost, and reliability for each transaction
  • Accelerate Settlement -- Automated reconciliation detects discrepancies and confirms settlement faster
  • Strengthen Compliance -- Systematic sanctions screening (OFAC, EU, UN) and AML checks on every payment
  • Improve Straight-Through Processing -- End-to-end automation reduces manual intervention

Architecture

graph TB
    Request["Client Request<br/><small>Payment Processing</small>"] --> Runtime["AgentCore Runtime"]
    Runtime --> Orchestrator["Orchestrator"]
    
    Orchestrator --> Validator["Payment Validator<br/><small>Compliance, sanctions, fraud checks</small>"]
    
    Validator --> Routing["Routing Agent<br/><small>Optimal payment rail selection</small>"]
    
    Routing --> Reconciliation["Reconciliation Agent<br/><small>Automated matching & settlement</small>"]
    
    Validator --> Bedrock["Amazon Bedrock<br/>(Claude)"]
    Routing --> Bedrock
    Reconciliation --> Bedrock
    
    Validator --> S3["S3 Sample Data<br/><small>data/samples/agentic_payments/</small>"]
    Routing --> S3
    Reconciliation --> S3
    
    Reconciliation --> Response["Response<br/><small>Validation, routing, reconciliation status</small>"]
    
    style Request fill:#e1f5ff
    style Response fill:#d4edda
    style Bedrock fill:#fff3cd
    style S3 fill:#fff3cd
Loading

Directory Structure

use_cases/agentic_payments/
├── README.md
└── src/
    ├── __init__.py                            # Framework router
    ├── strands/
    │   ├── __init__.py
    │   ├── config.py                          # Payment settings
    │   ├── models.py                          # PaymentRequest / PaymentResponse
    │   ├── orchestrator.py                    # AgenticPaymentsOrchestrator
    │   └── agents/
    │       ├── payment_validator.py           # PaymentValidator agent
    │       ├── routing_agent.py               # RoutingAgent agent
    │       └── reconciliation_agent.py        # ReconciliationAgent agent
    └── langchain_langgraph/                   # LangGraph implementation (same structure)

Agentic Design

The AgenticPaymentsOrchestrator extends StrandsOrchestrator and implements a parallel + conditional sequential pattern:

  1. Parallel Validation and Routing -- The Payment Validator and Routing Agent always run concurrently via asyncio.gather().
  2. Conditional Reconciliation -- For payment types requiring settlement tracking (wire, ACH, international), the Reconciliation Agent runs as a follow-up step.
  3. Synthesis -- A supervisor LLM call combines all agent outputs into a final recommendation (EXECUTE/REJECT/ESCALATE) with routing details and reconciliation status.

Agents

Payment Validator

Field Detail
Class PaymentValidator(StrandsAgent)
Role Validates payments against business rules, sanctions lists, and fraud indicators
Data Payment profile via s3_retriever_tool
Produces Validation status (APPROVED/REJECTED/REQUIRES_REVIEW), rules checked, violations, sanctions clearance, risk score (0-100)

Routing Agent

Field Detail
Class RoutingAgent(StrandsAgent)
Role Determines optimal payment rail based on cost, speed, and transaction requirements
Data Payment profile via s3_retriever_tool
Produces Selected rail (Fedwire/ACH/RTP/SWIFT/SEPA), alternative rails, estimated settlement time, routing cost, rationale

Reconciliation Agent

Field Detail
Class ReconciliationAgent(StrandsAgent)
Role Matches and reconciles payments across source and destination systems
Data Payment profile via s3_retriever_tool
Produces Reconciliation status (MATCHED/UNMATCHED/DISCREPANCY/PENDING), matched records, discrepancies, exception flags

Data and Tools

  • Tool: s3_retriever_tool -- Retrieves payment data from S3 by payment ID
  • S3 Path: data/samples/agentic_payments/{payment_id}/
  • Data Files: profile.json (payment details, amounts, parties, type)

Request / Response

Request (PaymentRequest)

class PaymentRequest(BaseModel):
    payment_id: str                            # e.g. "PMT001"
    payment_type: PaymentType                  # wire | ach | real_time | international | domestic
    additional_context: str | None = None

Response (PaymentResponse)

class PaymentResponse(BaseModel):
    payment_id: str
    transaction_id: str                        # UUID
    timestamp: datetime
    validation_result: ValidationResult | None # status, rules_checked, violations, sanctions_clear, risk_score
    routing_decision: RoutingDecision | None   # selected_rail, alternatives, settlement_time, cost, rationale
    reconciliation_status: ReconciliationStatus | None  # matched | unmatched | discrepancy | pending
    summary: str                               # Executive summary with EXECUTE/REJECT/ESCALATE
    raw_analysis: dict

Quick Start

# Deploy to AgentCore
USE_CASE_ID=agentic_payments ./scripts/deploy/full/deploy_agentcore.sh

# Test
./scripts/use_cases/agentic_payments/test/test_agentcore.sh

Sample Data

Payment ID Type Description
PMT001 Wire High-value domestic wire transfer ($125,000)

Related Documentation