CrewAI-based Credit Risk Demo with Plano Orchestration and Guardrails

[burney95]

Overview

This PR adds a complete, end-to-end demo that showcases how Plano can orchestrate a real multi-agent AI workflow in a practical financial use case.

The demo implements a credit risk assessment system built on top of CrewAI, routed entirely through Plano’s agent listener and LLM gateway. It highlights how multiple specialized agents can work together while Plano handles routing, model selection, security filters, and observability.

What’s Included

• A CrewAI-powered multi-agent workflow for credit risk assessment
  (intake → risk scoring → policy checks → decision memo)
• Integration with Plano’s agent listener and LLM gateway
• HTTP-based security filters for:
  • PII redaction
  • Prompt-injection detection
• A Streamlit UI to make the demo easy to run and understand
• OpenAI-compatible APIs preserved end-to-end
• Clear examples of how Plano enables:
  • Centralized orchestration
  • Guardrails and safety controls
  • Loose coupling between agent logic and infrastructure

Why This Demo Exists

The goal of this demo is to provide a concrete, runnable example that:

• Feels close to a real production workflow
• Shows how Plano fits naturally into an agentic system
• Demonstrates true multi-agent coordination, not just single-prompt usage

PII Redaction in Action

Prompt Injection Detection in Action