CrewAI

Role-based multi-agent orchestration. Define agents with roles, goals, and backstories; chain them into crews with sequential or hierarchical processes. CrewAI Flows add event-driven workflows; checkpoints let you resume long-running crews.

Install

pip install "crewai[tools]"

Version

v1.14.4 · May 2026 · Python 3.10+

Best for

Team simulations, research pipelines, content creation, analyst crews.

Where to start

Follow Zero → Hero for the tutorial path, or Jump to a topic if you already know what you need. The new API reference pages (Agent, Task, Memory, Knowledge, Tools, LLM, CLI) are verified against the installed source.

Zero → Hero

1 · Core fundamentals Install, LLM config, initialisation · 15 min

2 · Your first agent Role, goal, backstory, single-agent crews · 20 min

3 · Agent API deep dive Every field verified against source · 20 min

4 · Tasks & structured output Task API, guardrails, conditional tasks, processes · 25 min

5 · Tools & @tool decorator BaseTool, @tool, MCP tool bridging · 20 min

6 · Memory systems Unified Memory, scope/slice, storage backends · 20 min

7 · Knowledge sources PDF, CSV, Docling, string sources · 15 min

8 · LLM providers & routing Native SDK vs LiteLLM, reasoning/thinking · 15 min

9 · CLI & project layout crewai create/run/train/test/checkpoint · 15 min

10 · Flows (event-driven) @start/@listen decorators, state routing · 25 min

11 · Advanced multi-agent Hierarchies, delegation, collaboration · 30 min

12 · Production Deployment, monitoring, cost controls · 30 min

Jump to a topic

Agent fields & defaults

Task structured outputs

Process: sequential vs hierarchical

@tool decorator

BaseTool subclass

MCP server bridging

Memory scope & slice

Knowledge source types

LLM routing rules

Reasoning / thinking models

Custom BaseLLM

CLI cheat sheet

Project scaffold

Training & testing modes

Checkpoint CLI

CrewAI Flows

Reference

Agent API Fields, planning, guardrails, MCP, A2A.

Task API Fields, structured output, conditional tasks, processes.

Memory Unified Memory class, scope/slice, LanceDB/Qdrant.

Knowledge Built-in sources + custom BaseKnowledgeSource.

Tools BaseTool, @tool, MCP servers, caching, usage caps.

LLM providers Routing, native SDKs, LiteLLM fallback, reasoning.

CLI Every CLI command verified against source.

Comprehensive guide Single-page technical tour (some errata — see note).

Advanced multi-agent

Flows guide Event-driven workflows.

Recipes Blog generation, BI, support, code review, marketing, legal.

Diagrams

Production guide

What’s new in v1.14.4 (May 2026)

• v1.14.4 (May 1) — Patch release; stability improvements.

What’s new in v1.14.3 (April 2026)

• New reference pages — Agent, Task, Memory, Knowledge, Tools, LLM, CLI — each verified against installed source.
• Checkpoint forking with lineage tracking — branch execution from any saved checkpoint.
• from_checkpoint on Crew.kickoff / Flow.kickoff — resume a crew or flow with from_checkpoint=CheckpointConfig(restore_from=...).
• Checkpoint CLI — crewai checkpoint list|info|resume|diff|prune.
• Template management CLI — crewai template list|add.
• Enriched token tracking — reasoning_tokens, cache_creation_tokens on CrewOutput.token_usage.
• Scoped streaming handlers — per-run isolation across concurrent crews.

Breaking changes since v1.5.0

• CodeInterpreterTool removed from the public API — use E2B / Modal / a custom BaseTool.
• CrewStructuredTool moved to internal API — external code should use BaseTool.
• stop= parameter stripped for GPT-5 / o-series models by the LLM router.
• Agent.reasoning deprecated → use planning + PlanningConfig.
• Agent.multimodal deprecated → use Task.input_files={"name": path}.
• from crewai.amp import ... is not a valid import in the stable package. AMP is accessed through the CLI (crewai login, crewai deploy ...). The AMP Suite section of the comprehensive guide includes pre-existing fabricated imports and is flagged with an errata note.
• v1.10.0 yanked — use 1.10.1+.

Install, minimal example

from crewai import Agent, Task, Crew, Process, LLM

llm = LLM(model="openai/gpt-4o-mini")

analyst = Agent(
    role="Research Analyst",
    goal="Analyse topics thoroughly",
    backstory="Expert analyst with 10 years of experience",
    llm=llm,
)

task = Task(
    description="Research AI advancements in 2026",
    expected_output="Concise analysis of key developments",
    agent=analyst,
)

crew = Crew(agents=[analyst], tasks=[task], process=Process.sequential)
result = crew.kickoff()
print(result)

Ready? Start Zero → Hero Step 1 →

---

Revision history

Date	Version	Changes	Reviewer
2026-05-01	1.14.4	Version bumped 1.14.3 → 1.14.4; “What’s new” section updated; revision history entry added. Agent, Task, Crew imports verified against installed 1.14.4 (.routine-envs/check-crewai-0501).	Claude routine
2026-04-25	1.14.3	Pre-release 1.14.3a2 confirmed stable as v1.14.3; version card updated from pre-release to stable; revision history entry added. Symbols verified against installed crewai 1.14.3 (.routine-envs/crewai-py-0425).	Claude routine
2026-04-22	1.14.3a2	Added 7 source-verified reference pages: Agent API, Task API, Memory, Knowledge, Tools, LLM providers, CLI. Corrected checkpoint import path (crewai.state.provider.sqlite_provider), field name (checkpoint=, not checkpoint_config=), and resume API (from_checkpoint=CheckpointConfig(restore_from=...)). Added errata banner to the AMP Suite section (imports fabricated; use CLI).	Claude (Opus 4.7)
2026-04-21	1.14.2	Index redesigned with Zero → Hero + Jump-to-topic grid.	—
April 17, 2026	1.14.2	Checkpoint forking; from_checkpoint; checkpoint CLI; template CLI; reasoning_tokens.	—
April 7, 2026	1.14.0	Checkpoint system; Pydantic structured outputs; SSRF protections; RBAC; MCP custom servers.	—
November 2025	1.5.0	CrewAI Flows; MCP integration; memory systems; hierarchical processes.	—