Product Lifecycle Agent

Product Lifecycle Agent is an AI-assisted operations workflow for identifying underperforming SKUs, proposing commercial actions, validating feasibility, and producing transparent decision reports.

The system uses a sequential multi-agent pipeline with human-in-the-loop governance. It is built to show how agentic workflows can support product, pricing, marketing, labor, finance, and governance decisions from structured business data.

Highlights

• Seven-stage agent workflow for product lifecycle decisions
• Deterministic business-rule agents for pricing, labor, finance, product, and governance checks
• LLM-powered report generation for human-readable transparency reports
• Flask web app for CSV upload, configuration, workflow execution, approvals, and downloads
• MCP server layer for centralized data access and future database-backed execution
• Audit-oriented outputs for every stage of the decision process
• Command-line workflow runner for repeatable tests

Agent Workflow

TriggerAgent
  -> PricingAgent
  -> MarketingAgent
  -> LaborAgent
  -> FinanceAgent
  -> ProductAgent
  -> GovernanceAgent
  -> ReportAgent

Each agent writes its output to disk so the full reasoning path can be inspected, reviewed, and audited.

Tech Stack

• Python
• Flask
• CrewAI
• MCP
• Pandas
• Pydantic
• OpenAI
• CSV-based data layer with MCP migration path

Project Structure

agents/          trigger, pricing, marketing, labor, finance, product, governance, report agents
tools/           shared data access helpers
mcp_servers/     MCP server and data functions
webapp/          Flask UI for uploads, workflow execution, and governance review
data/input/      source CSV files
data/output/     generated workflow results
crew.py          workflow orchestration
test_workflow.py command-line runner

Local Development

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Run the web app:

python run_webapp.py

Run the workflow from the command line:

python test_workflow.py

Optional MCP mode:

./start_mcp_server.sh
python test_workflow.py
./stop_mcp_servers.sh

Input Files

Place CSV inputs in data/input/:

• get_sales.csv
• get_inventory.csv
• price_optimizer.csv

Security

Do not commit .env, logs, generated output folders, customer data, or real operational exports.