Multi-AI Agent Systems with CrewAI 🤖

Beyond single agents — real-world automation with autonomous multi-agent teams.
A production-oriented collection of multi-agent systems covering sequential, parallel, and hierarchical orchestration patterns for business automation.

---

Table of Contents

• Overview
• Architecture
• Project Structure
• Use Cases
• Core Skills
• Getting Started
• Tech Stack

---

🔍 Overview

Single agents hit a ceiling. Complex real-world tasks require specialized agents working in parallel, supervised by a manager, passing results to one another — just like a human team.

This repository implements six autonomous multi-agent systems, each targeting a distinct business problem: from content pipelines and customer support to financial analysis and recruitment automation. Every system is built with CrewAI and demonstrates a different orchestration pattern that scales from prototype to production.

---

🏗 Architecture

                    ┌──────────────────────┐
                    │     Manager LLM      │  ← Hierarchical Mode
                    └──────────┬───────────┘
                               │ coordinates
              ┌────────────────┼────────────────┐
              ▼                ▼                ▼
       ┌────────────┐  ┌────────────┐  ┌────────────┐
       │  Agent A   │  │  Agent B   │  │  Agent C   │
       │ (Planner)  │  │ (Analyst)  │  │  (Writer)  │
       └─────┬──────┘  └─────┬──────┘  └─────┬──────┘
             │               │               │
             ▼               ▼               ▼
          Tools           Tools           Tools
      (SerperDev)       (Custom)      (MDXSearch)
             │               │               │
             └───────────────┴───────────────┘
                             │
                     ┌───────▼────────┐
                     │ Pydantic Output │  ← Validated JSON
                     └────────────────┘

---

📁 Project Structure

multi-agent-systems-crewai/
│
├── 01_Research_and_Write_Article/     # Sequential: Planner → Writer → Editor
├── 02_Customer_Support_Automation/    # Role-playing agents with shared memory
├── 03_Customer_Outreach_Campaign/     # Custom tools + sentiment analysis
├── 04_Automated_Event_Planning/       # Async parallel tasks + human gates
├── 05_Financial_Analysis_System/      # Hierarchical: Manager LLM + specialists
├── 06_Tailor_Job_Applications/        # RAG pipeline + multi-tool research
│
├── .env.example
├── requirements.txt
└── README.md

---

🎯 Use Cases

01 · Research & Write Article

Pattern: Sequential
A three-agent pipeline where a Planner researches the topic, a Writer drafts the article, and an Editor refines the final output. Clean separation of responsibilities with task handoff.

02 · Customer Support Automation

Pattern: Role-Playing + Memory
Agents that retain conversation context across interactions using built-in memory management. Demonstrates how multi-agent cooperation improves response quality in support scenarios.

03 · Customer Outreach Campaign

Pattern: Custom Tool Integration
Marketing automation pipeline with a custom Sentiment Analysis tool and directory-based content reading. Shows how to extend CrewAI agents with bespoke Python tooling.

04 · Automated Event Planning

Pattern: Async Parallel + Human-in-the-Loop
Tasks run in parallel (async) across multiple agents simultaneously. Strategic interrupt gates pause execution for human approval before committing actions. Uses Pydantic for structured, validated JSON outputs.

05 · Financial Analysis System

Pattern: Hierarchical Orchestration
A Manager LLM dynamically assigns tasks to specialized analyst and risk agents. The most advanced orchestration pattern — closest to how real enterprise AI pipelines operate.

06 · Tailor Job Applications

Pattern: RAG + Multi-Tool Research
A full recruitment automation pipeline using MDXSearchTool for semantic search over local documents, combined with web research tools. Produces tailored application materials automatically.

---

🛠 Core Skills Demonstrated

Skill	Implementation
Agent Orchestration	Sequential, Parallel (Async), Hierarchical flows
Custom Tool Engineering	Python-native tools: Sentiment Analysis, directory readers
Structured Output	Pydantic-validated JSON responses across all agents
Human-Agent Collaboration	interrupt_before approval gates in critical workflows
RAG & Semantic Search	MDXSearchTool for local document querying
Memory Management	Cross-turn context retention for support agents

---

🚀 Getting Started

Prerequisites

• Python 3.10+
• OpenAI API key
• Serper API key (free tier at serper.dev)

Installation

# 1. Clone the repository
git clone https://github.com/ozereray/multi-agent-systems-crewai.git
cd multi-agent-systems-crewai

# 2. Install dependencies
pip install -r requirements.txt

# 3. Configure environment
cp .env.example .env

Configuration

OPENAI_API_KEY=your_openai_api_key
SERPER_API_KEY=your_serper_dev_api_key

Run Any System

# Sequential article writer
python 01_Research_and_Write_Article/main.py

# Hierarchical financial analysis (most advanced)
python 05_Financial_Analysis_System/main.py

# Recruitment pipeline with RAG
python 06_Tailor_Job_Applications/main.py

---

🛠 Tech Stack

Tool	Role
CrewAI	Multi-agent orchestration framework
LangChain	LLM abstraction & tool layer
OpenAI GPT-4o	Core reasoning model
SerperDevTool	Real-time web search for agents
Pydantic	Structured & validated agent outputs

---

🤝 Contributing

Issues and pull requests are welcome. Each use case is self-contained — contributions to individual systems are encouraged.

---

Built with 🤖 for real-world agent automation
github.com/ozereray