CrewAI in Action

Building Production-Ready AI Agents and Multi-Agent Systems

Introduction

1. The Problem with Single-Agent Systems
2. The Multi-Agent Solution
3. Why CrewAI
4. What This Book Covers
5. How to Use This Book
6. What You Will Need
7. A Note on Versions

Chapter 1: Getting Started with CrewAI

1. Why Multi-Agent Systems
2. Installing CrewAI and Dependencies
3. Your First CrewAI Project
4. Understanding the Project Structure
5. Running Your First Crew
6. Common First-Run Issues
7. Next Steps

Chapter 2: Agents, Tasks and Crews

1. Designing Effective Agents
2. Crafting Precise Tasks
3. Orchestrating Crews
4. Sequential vs Hierarchical Processes
5. Structured Outputs with Pydantic
6. Putting It All Together
7. Common Pitfalls and Troubleshooting
8. Summary

Chapter 3: Connecting to Language Models

1. The LLM Layer in CrewAI
2. OpenAI Integration
3. Anthropic Claude Integration
4. Google Gemini and Other Providers
5. Local Models with Ollama
6. Per-Agent Model Selection
7. Model Selection Guidelines
8. Configuring LLM Parameters
9. Provider Configuration Examples
10. Handling Provider Failures
11. Summary

Chapter 4: Tools and Tool Integration

1. Understanding the Tool System
2. Built-In CrewAI Tools
3. Creating Custom Tools with the Decorator
4. Class-Based Custom Tools
5. Tool Input Validation with Pydantic
6. Asynchronous Tool Implementations
7. Tool Error Handling
8. Tool Organization Patterns
9. Summary

Chapter 5: Model Context Protocol (MCP) Integration

1. What Is the Model Context Protocol
2. Adding MCP Servers as Tools
3. Popular MCP Servers
4. Building Your Own MCP Server
5. SSE Transport for Remote Servers
6. MCP Security Considerations
7. Real-World MCP Integration Patterns
8. Summary

Chapter 6: Memory Systems

1. The Need for Agent Memory
2. Short-Term and Long-Term Memory
3. Entity Memory and Knowledge Graphs
4. The Unified Cognitive Memory System
5. External Memory with Vector Databases
6. Memory in Practice: A Research Assistant
7. Memory Pitfalls and Best Practices
8. Summary

Chapter 7: Knowledge Integration and RAG

1. Agent Knowledge Bases
2. Building RAG Pipelines with CrewAI
3. Vector Database Integration
4. The RAG Tool and Embedding Providers
5. Multi-Agent RAG Systems
6. RAG Optimization Strategies
7. Common RAG Pitfalls
8. Summary

Chapter 8: Flows and Event-Driven Orchestration

1. From Crews to Flows
2. Flow State Management
3. The Start, Listen and Router Decorators
4. Conditional Logic and Branching
5. Checkpointing and Recovery
6. Streaming and Real-Time Output
7. Flow Visualization
8. Building Flows in Practice
9. Summary

Chapter 9: Guardrails, Validation and Safety

1. Why Guardrails Matter
2. Function-Based Guardrails
3. LLM-Based Guardrails
4. Hallucination Detection
5. Input Validation and Sanitization
6. Retry Configuration
7. Guardrails in Production Workflows
8. Summary

Chapter 10: Building Real-World Applications

1. Multi-Agent Research Assistant
2. Customer Support Automation System
3. AI Coding Assistant with Code Review
4. Document Analysis and Summarization Pipeline
5. Summary

Chapter 11: Asynchronous Workflows and Performance Optimization

1. Async Task Execution
2. Parallel Crew Execution
3. Token Usage Optimization
4. Cost Reduction Strategies
5. Caching and Semantic Caching
6. Performance Monitoring
7. Summary

Chapter 12: Testing, Debugging and Observability

1. Debugging Agent Behavior
2. Structured Logging Patterns
3. Tracing and Monitoring
4. Testing Strategies for Agent Systems
5. Third-Party Observability Integrations
6. Summary

Chapter 13: Production Deployment and Scaling

1. Containerizing CrewAI Applications
2. Building REST APIs Around Crews
3. Security Best Practices
4. Deployment Platforms and Strategies
5. Scaling and High Availability
6. Monitoring in Production
7. Summary

Conclusion

References