Multi-agent AI applications for financial services. This directory contains three categories of applications, each serving a different adoption path.
- Reference Implementations
- FSI Foundry
- App Factory
- Which Should I Use?
- Getting Started
- Directory Structure
- Documentation
End-to-end full-stack solutions for specific FSI use cases. Each implementation is a complete, self-contained project with its own frontend, backend, infrastructure, and deployment pipeline.
| Implementation | Domain | Description |
|---|---|---|
| Trade Surveillance | Capital Markets | Real-time trade monitoring, pattern detection, and regulatory alerting |
| Intelligent Document Processing | Operations | Automated document ingestion, classification, extraction, and validation |
View all reference implementations
A collection of multi-agent POC implementations spanning banking, payments, risk & compliance, capital markets, insurance, and operations — all built on one shared foundation of infrastructure and backend code.
- One codebase, three deployment patterns — EC2/ALB, Step Functions, or AgentCore Runtime
- Two framework implementations per use case — LangGraph/LangChain and Strands
- Shared foundations — adapters, base classes, Terraform modules, Docker configs, agent registry
- Deploy any use case with a single interactive script
34 use cases across 7 FSI domains:
| Domain | Count | Examples |
|---|---|---|
| Banking | 8 | KYC Risk Assessment, Customer Service, Customer Chatbot, Corporate Sales |
| Payments | 3 | Agentic Payments, Payment Operations, Fraud Detection |
| Risk and Compliance | 5 | Credit Risk Assessment, Compliance Investigation, Market Surveillance |
| Capital Markets | 9 | Investment Advisory, Trading Assistant, Earnings Summarization |
| Insurance | 3 | Customer Engagement, Claims Management, Life Insurance Agent |
| Operations | 3 | Call Center Analytics, Post Call Analytics, Call Summarization |
| Modernization | 3 | Legacy Migration, Code Generation, Mainframe Migration |
View FSI Foundry documentation
Declarative markdown blueprints that describe complete agentic applications end-to-end — agent logic, infrastructure, deployment pipelines, and tests — designed to be fed into AI coding assistants to generate fully functional apps.
| Scenario | Recommended Path |
|---|---|
| Need a complete, deployable solution for a specific use case | Reference Implementations |
| Exploring multi-agent patterns across FSI domains | FSI Foundry |
| Building multiple use cases on shared infrastructure | FSI Foundry |
| Generating a new application from a specification | App Factory |
# Deploy an FSI Foundry use case (interactive)
cd fsi_foundry
./scripts/main/deploy.sh
# Explore a reference implementation
cd reference_implementations/<project-name>
# Follow the project-specific READMEapplications/
├── reference_implementations/ # End-to-end full-stack solutions
├── fsi_foundry/ # Multi-agent POCs on shared foundations
│ ├── foundations/ # Reusable infrastructure
│ │ ├── src/ # Adapters, base classes, tools
│ │ ├── iac/ # Terraform modules (EC2, SF, AgentCore)
│ │ ├── docker/ # Container configurations
│ │ └── ui/ # Testing dashboard
│ ├── use_cases/ # 34 FSI use case implementations
│ ├── data/ # Agent registry and sample data
│ ├── docs/ # Architecture, deployment, security
│ └── scripts/ # deploy.sh, test.sh, cleanup.sh
├── app_factory/ # Markdown blueprints for app generation
└── README.md # This file
| Resource | Description |
|---|---|
| Reference Implementations | Available full-stack solutions |
| FSI Foundry | Architecture, foundations, and use cases |
| Deployment Guide | Step-by-step deployment instructions |
| Security | Threat model, AI security, AWS service security |
| Adding Use Cases | How to add new use cases to FSI Foundry |
| App Factory | Blueprint specifications |
See LICENSE for details.