CrewAI Template

Defines crews as collections of specialized agents with distinct roles, goals, and backstories, orchestrated through a Crew → Agent → Task hierarchy. Agents are configured via YAML files (e.g., gamedesign.yaml) that specify agent personality, tools, and task dependencies, enabling declarative composition of multi-agent workflows without code changes. The framework handles agent sequencing, context passing between agents, and collaborative task execution through a centralized crew coordinator.

Implements advanced workflow orchestration using CrewAI Flow framework with state machines, conditional branching, and asynchronous processing. Workflows support human oversight checkpoints (e.g., Lead Score Flow approval gates) where humans review and approve agent decisions before proceeding. The flow system manages complex state transitions, parallel task execution, and interactive decision routing based on agent outputs, enabling workflows like lead scoring with approval, email auto-response, and book writing with chapter reviews.

Implements a structured book writing system using CrewAI Flow where agents collaborate on chapter composition, outline generation, and content review. The Book Writing Flow demonstrates how agents can work sequentially on different chapters, maintain narrative consistency, and incorporate human feedback at review checkpoints. The workflow manages chapter dependencies, ensures thematic coherence, and allows human editors to approve or request revisions before proceeding to the next chapter.

Implements a lead qualification system using CrewAI Flow that analyzes lead data, scores prospects based on predefined criteria, and routes high-value leads through approval gates before action. The Lead Score Flow demonstrates conditional routing where leads above a score threshold proceed to different agents than lower-scoring leads. Human reviewers can approve or reject scoring decisions, and the workflow generates personalized follow-up actions based on lead quality and approval status.

Implements a meeting support system using CrewAI Flow that processes meeting transcripts or notes, generates structured agendas, extracts action items, and identifies key decisions. The Meeting Assistant Flow demonstrates how agents can analyze unstructured meeting content, identify participants, extract decisions and commitments, and generate follow-up action items with ownership. The workflow supports both pre-meeting agenda generation and post-meeting analysis.

Implements an automated landing page creation system using CrewAI where agents collaborate on copywriting, design specification, and layout composition. The Landing Page Generation Flow demonstrates how agents can generate marketing copy, define page structure, select design templates, and compose HTML/CSS based on specifications. The workflow supports A/B testing variants and enables rapid iteration on landing page designs.

Executes tasks in a defined sequence where each agent receives the output of previous agents as context, enabling knowledge accumulation across the workflow. The framework maintains a shared context object that flows through the agent chain (e.g., Game Builder Crew: game concept → design document → implementation plan). Each task's output becomes input to the next task, with the crew coordinator managing context passing, preventing information loss, and ensuring agents build on prior work rather than starting from scratch.

Supports multiple LLM providers (OpenAI, Anthropic, NVIDIA NIM, Azure OpenAI) through a unified agent interface, allowing agents to be configured with different models without code changes. The framework abstracts provider-specific APIs, enabling seamless switching between local models (Ollama), proprietary APIs (OpenAI), and enterprise solutions (Azure). Configuration specifies the LLM provider per agent, enabling heterogeneous crews where different agents use different models based on task requirements and cost optimization.

Extends agent capabilities by binding external tools (APIs, functions, web search) to agents through a tool registry. Agents can invoke tools during task execution to fetch real-time data, perform calculations, or interact with external systems. The framework handles tool invocation, parameter binding, and result integration back into agent reasoning. Examples demonstrate tools for web search (trip planning), financial data retrieval (stock analysis), and API calls (weather, LinkedIn integration for recruitment).

Provides pre-built crew templates optimized for specific domains: Game Builder (sequential game development), Stock Analysis (financial research with SEC filings), Marketing Strategy (content generation with NVIDIA AI), Job Posting (automated job description creation), Recruitment (LinkedIn integration and candidate scoring), Trip Planning (travel itinerary with weather/search APIs), and Book Writing (multi-chapter composition with reviews). Each template includes pre-configured agents, tasks, tools, and YAML configurations that can be adapted for similar use cases without building crews from scratch.

Enables declarative definition of agents and tasks through YAML configuration files (e.g., gamedesign.yaml) rather than Python code. YAML files specify agent roles, goals, backstories, tools, and task descriptions with dependencies. The framework parses YAML and instantiates agents and tasks at runtime, enabling non-developers to modify agent behavior, add tasks, or adjust tool bindings without touching Python. Configuration changes are immediately reflected in crew execution without recompilation.

Supports asynchronous task execution within flows, enabling parallel processing of independent tasks and non-blocking workflow progression. The Flow framework manages async task scheduling, concurrent execution, and result aggregation. Tasks can run in parallel when they have no dependencies, reducing total workflow execution time. Examples demonstrate async patterns in Lead Score Flow and Email Auto-Responder Flow where multiple agents process data concurrently.

Provides integration patterns with LangGraph for advanced workflow orchestration, enabling CrewAI crews to be embedded within LangGraph state machines. The CrewAI-LangGraph example demonstrates how to compose crews as nodes in a LangGraph workflow, combining CrewAI's agent orchestration with LangGraph's graph-based state management and visualization. This enables hybrid workflows where CrewAI handles agent coordination and LangGraph manages overall workflow topology.

Implements an automated email response system using CrewAI Flow that analyzes incoming emails, classifies them, and generates contextually appropriate responses using templates. The Email Auto-Responder Flow demonstrates how agents can process unstructured email content, extract key information, route to appropriate response agents, and generate personalized replies. The workflow includes conditional routing based on email classification and template selection based on email type.