Skip to content
Get packs

Guide · Agent skill

How to use the CrewAI Multi-Agent Orchestration agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The crewai-multi-agent skill is a set of instructions and reference files that teaches an AI agent how to build CrewAI projects: teams of agents with defined roles, tasks and a run order. It is aimed at Python developers who want to prototype multi-agent workflows. After this guide you will know how to install the skill, how to ask for it in plain language, and what you need before running any generated code.

CrewAI Multi-Agent OrchestrationOrchestra-Research/AI-Research-SKILLs · MIT

Quick start

  1. 1Download and unzip the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
  2. 2Install Python packages. Install crewai and crewai[tools] with pip so the versions meet the 1.2.0 minimum.
  3. 3Add your API keys. Provide an LLM API key such as OpenAI's. Add a search key only if you plan to use SerperDevTool.
  4. 4Ask for a crew in plain language. Describe the team roles, the tasks and the order they should run in.
  5. 5Review the generated code. Check which tools, files and network calls the code uses before running it.
  6. 6Run the project. Use crewai install and crewai run inside the generated project folder.

What the skill makes the agent do

Once the skill is loaded, the agent knows how to set up a CrewAI project. It defines several agents, each with a role, a goal and a backstory, then assigns them tasks. The tasks are run as a crew, either one after another (sequential) or with an automatically created manager agent that delegates work (hierarchical).

The skill also covers connecting built-in tools such as web search, web scraping and file reading, writing custom tools, and using memory and event-driven Flows. Its reference files add guidance on flows, tools and troubleshooting.

  • Define agents with role, goal, backstory and optional tools
  • Chain tasks with context so later tasks use earlier output
  • Choose sequential or hierarchical process execution
  • Add built-in tools (search, scraping, file reading) or write custom ones

Install the skill from the ZIP

Download the ZIP for this skill from this page and unzip it. You can put the skill folder in one of two places: your personal skills folder at ~/.claude/skills/, or a project-level .claude/skills/ folder inside a specific project. A personal install is available in every project; a project install applies only to that project.

If you use the Claude apps instead, you can upload the skill under Settings → Capabilities → Skills.

  • Personal install: ~/.claude/skills/
  • Project install: .claude/skills/ inside your project
  • Claude apps: Settings → Capabilities → Skills, then upload

Set up the Python packages the skill needs

The generated code needs Python, the crewai package (version 1.2.0 or newer) and the crewai-tools package (also 1.2.0 or newer). The skill's reference notes say CrewAI requires Python 3.10 to 3.13. Install the core framework with the tools extra so the built-in tools are available.

You also need an LLM API key, such as OpenAI's; the examples use gpt-4o. If you want the SerperDevTool search tool, you need a separate search API key.

  • Python 3.10 to 3.13
  • crewai>=1.2.0 and crewai-tools>=1.2.0
  • An LLM API key such as OpenAI (examples use gpt-4o)
  • Optional: a search API key for SerperDevTool (the reference names SERPER_API_KEY)
pip install crewai pip install 'crewai[tools]'

Trigger the skill with natural requests

You do not need to name a command to use the skill. Describe the multi-agent workflow you want in plain language and mention the kind of team, the task and the order of work. The skill's description covers requests about teams of specialized agents, role-based collaboration with memory, and sequential or hierarchical pipelines.

Good requests name the roles and the output you expect. Vague requests tend to produce generic agents, so say what each agent should own.

  • "Build a CrewAI crew with a researcher and a writer that produces a 500-word blog post on a topic I give."
  • "Set up a hierarchical CrewAI process where a manager delegates research, writing and review tasks."
  • "Create a CrewAI project with a custom calculator tool and add it to my analyst agent."
  • "Show me how to add memory to my CrewAI crew and when to use a Flow instead of a Crew."

Run the example crew

The skill's quick start creates a project with the CrewAI command line tool, installs its dependencies and runs it. You can also write a crew directly in code: define agents, define tasks, create a Crew with a process, then call kickoff with inputs. The result's raw attribute holds the final output.

Expect the agent to ask you to run these commands itself or to show them to you. Read the generated code before running it, because it will call your LLM provider and may use tools.

  • Run crewai create crew my_project, then cd into it
  • Run crewai install and crewai run
crewai create crew my_project cd my_project crewai install crewai run

Choose sequential, hierarchical or Flows

Use a sequential process when the steps are fixed and each one feeds the next. Use a hierarchical process when you want a manager agent to decide how work is split. Use Flows when you need conditional branching, explicit state or event-driven steps; the reference table suggests Crews for simple collaboration and Flows for complex state and branching.

  • Sequential: fixed order, default in the examples
  • Hierarchical: auto-created manager agent delegates tasks; set manager_llm
  • Flows: conditional branching, complex state, event-driven workflows, or crews inside flow steps

What people use it for

  • Prototyping a researcher-and-writer pipeline that produces a draft report from a topic.
  • Building a hierarchical crew where a manager agent splits research, writing and review work.
  • Adding a custom calculator tool to an analyst agent for a data question.
  • Comparing a CrewAI workflow with an equivalent LangChain or LangGraph design.

Frequently asked questions

How do I install the crewai-multi-agent skill?

Download the ZIP from this page and unzip it into ~/.claude/skills/ for a personal install, or into a project's .claude/skills/ folder. In the Claude apps you can instead upload it under Settings → Capabilities → Skills.

What do I need to run CrewAI code from this skill?

You need Python, the crewai and crewai-tools packages at version 1.2.0 or newer, and an LLM API key such as OpenAI's. A search API key is only needed for SerperDevTool.

Is the crewai-multi-agent skill free?

The skill is published under the MIT licence. The LLM APIs it calls are paid services, and the skill does not enforce rate limits or spending caps, so check your provider's pricing before running crews.

Does CrewAI need LangChain installed?

No. The framework is described as standalone with no LangChain dependencies, which keeps its footprint small.

How does CrewAI compare with LangChain or LangGraph?

The skill positions CrewAI for role-based multi-agent teams and says it is simpler to set up than LangChain or LangGraph. It suggests LangChain for general LLM apps and RAG pipelines, and LangGraph for complex stateful workflows with cycles.

Why does my CrewAI agent keep looping?

The troubleshooting reference suggests setting max_iter to limit iterations and max_rpm to rate-limit calls, and writing a clearer task description. A ModuleNotFoundError for crewai_tools is usually fixed with pip install 'crewai[tools]'.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits Python developers who want an agent to scaffold CrewAI teams, compare sequential and hierarchical processes, or test role-based workflows. It is not suited to non-programmers or to sensitive data, because the code can read files, fetch web pages and make paid LLM calls that you should review first. If you want a general LLM app or retrieval pipeline rather than a multi-agent team, consider LangChain or LlamaIndex instead.

Download the skill