CrewAI is a framework for building multi-agent applications around specialized roles, autonomous agents, tasks, memory, and tools. It favors structured, enterprise-oriented coordination over open-ended agent interaction. section 4.0 section 4.4
Build a crew by combining these parts:
- Agents define a role, goal, verbosity, memory, backstory, delegation permissions, and optional tools. section 4.4.1
- Tasks describe the work, expected output, assigned agent, and execution behavior. section 4.4.1
- A Crew connects agents and tasks, then configures process type, memory, caching, request limits, and sharing. section 4.4.1
- Sequential processing passes work through the listed agents and tasks; hierarchical processing coordinates them through a crew manager. section 4.4 section 4.5
From jokes to software teams
A joke crew assigns research to a senior joke researcher and writing to a joke writer, then starts the workflow with a topic such as “AI engineer jokes.” A coding crew instead combines a senior software engineer, QA engineer, and chief QA engineer to create, inspect, and evaluate a Python game. Switching that crew from sequential to hierarchical processing adds a manager model for coordination. section 4.4.1 section 4.5
Figure 4.10 The composition of a CrewAI system
Use the system composition diagram to trace how agents, tasks, tools, memory, and processing patterns connect inside a CrewAI crew. section 4.4
