CrewAI Examples

CrewAI is a framework for building multi-agent AI systems where specialized agents collaborate to complete complex tasks. AgentVisor integrates natively with CrewAI — your crews run inside a hardened sandbox with policy enforcement, credential brokering, and durable execution via Temporal.

Prerequisites

• Anthropic API key (or another LLM endpoint) — the examples use Claude via the AgentVisor proxy
• Temporal — required for durable execution
• AgentVisor CLI — agentvisor serve . starts the runtime

Examples

Example	Difficulty	Key Feature
Research Agent	Intermediate	Multi-agent crew with MCP web fetch tools

How It Works

CrewAI agents in AgentVisor work through the CrewAI framework provider:

1. Discovery: AgentVisor detects crewai.yaml and selects the CrewAI provider automatically
2. Schema extraction: Agent input/output schemas are extracted from task {placeholder} tokens
3. Execution: python -m agentvisor.crewai.runner loads and runs your crew inside the sandbox
4. I/O: Input JSON is written to the runner's stdin; output JSON is read from stdout

Key Concepts

Concept	What It Is	Where You'll See It
crewai.yaml	Discovery file pointing AgentVisor to your crew classes	All examples
@CrewBase	Decorator for declarative crew definition	All examples
@agent, @task, @crew	CrewAI declarative decorators	All examples
get_mcp_tools()	AgentVisor SDK function to load MCP tools for agents	Research Agent
{placeholder}	Input variable syntax in task descriptions	All examples

For a deep dive into CrewAI integration — including crewai.yaml format, schema discovery, and provider options — see the CrewAI Agents Guide.

Quick Start

# Create a project from the research agent template
agentvisor template create crewai/research-agent
cd research-agent

# Run with Temporal started separately
temporal server start-dev &
agentvisor serve . --sandbox=none