CrewAI CLI Cheatsheet - crewai Multi-agent Framework Full Reference
CrewAI splits multi-agent collaboration into two shapes that differ not in capability but in determinism. A crew hands a set of roles and tasks to the framework to orchestrate; a flow is code you write yourself, defining steps and branches. The first saves you code at the cost of an execution path that is not guaranteed identical between runs; the second costs a little more code and buys reproducibility, debuggability and breakpoints. Which one to pick depends on whether the process itself must be controlled: content production and research summaries, where a correct result is enough, usually suit a crew, while anything involving approval, fallback and retry generally needs a flow. This sheet covers install and creation, running and training, tools and maintenance, and help and troubleshooting, noting that create crew and create flow generate two entirely different directory layouts that should not be mixed. Reach for it when behaviour looks wrong: first confirm which shape you are running, then consider reset-mem, because a stale memory file is the real reason behind tasks edits that appear to do nothing.
Install and Create 4
pip install crewaicrewai create crew my_crewcrewai create flow my_flowcrewai --helpRun and Train 6
crewai runcrewai traincrewai flow kickoffcrewai flow plotcrewai installcrewai run --helpTooling and Maintenance 8
crewai tool create my_toolcrewai tool listcrewai tool --helpcrewai reset-memcrewai versioncrewai create crew --helpcrewai train --helpcrewai flow --helpHelp and Troubleshoot 3
crewai create flow --helpcrewai tool create --helpcrewai install --helpTips
- crewai create crew generates a runnable scaffold; edit agents and tasks before run.
- flow is lighter than crew and more deterministic; use flow for simple pipelines.
- reset-mem helps when behavior is odd, clearing stale memory interference.
FAQ
How is CrewAI different from LangGraph?
CrewAI centers on role-based multi-agent collaboration with clear conventions; LangGraph orchestrates arbitrary state machines via graphs, more flexible but more boilerplate.
Must I write Python?
Core agents and tasks are defined in Python, but the crewai CLI handles most engineering tasks like create, run, and train.
Official References
Each command links to its official documentation below, so you can verify the latest usage and read deeper.
Maintained by LaoHand
Publicly updated on Sep 10, 2026, continuously proofread against official docs.
Contact Us
Wrong command or description? Send us corrections, business inquiries or product feedback by email.
Contact Us