Module DD-12 — CrewAI: Role-Based Multi-Agent (Emergent Coordination)

CrewAI: Role-Based Multi-Agent (Emergent Coordination)

45,900+ stars. Most accessible multi-agent framework. Role-based sequential/concurrent crews. 1.8s avg latency. The emergent-coordination pole — paired with LangGraph (DD-10) as the declared-coordination pole, both load-bearing for Course 4 E09 (Multi-Agent Orchestration).

minutes

artifacts

sub-sections

CrewAI optimizes for accessibility — the fastest way to stand up a multi-agent crew with role-based task assignment, and the most legible mental model (describe the team) in the multi-agent category. The coordination graph emerges from task context references and the process type, not from explicit drawn edges. This is the emergent-coordination pole of the multi-agent axis; LangGraph (DD-10) is the declared-coordination pole. The contrast is load-bearing for Course 4 E09. CrewAI sacrifices production-readiness: no sandboxing, crew-scoped shared memory with no write isolation, limited state, and run logs below the structured-event floor.

Key Claims

Load-Bearing Claims

CrewAI and LangGraph (DD-10) define the multi-agent coordination axis: emergent vs declared. CrewAI is the emergent-coordination pole — define roles and tasks, the graph emerges from context refs and process type. LangGraph is the declared-coordination pole — draw nodes and edges, the graph is a first-class object. Every multi-agent framework sits between these two poles. This axis is the organizing principle of Course 4 E09.

Role is the system prompt — the most accessible multi-agent mental model in the roster. CrewAI does not give you a raw system-prompt string; it gives you a templated one whose fields are role, goal, backstory. Module 12 (Prompt Assembly) applied at agent granularity. 'Describe the team' beats 'draw the graph' for first-time builders. The cost: when the template is wrong, you fight the framework's opinion.

Sequential crews earn the 4/5 on Module 1.3 (Subagents) — a textbook multi-agent pattern. Tasks run in list order; each output chains forward as context for the next agent. A linear ReAct chain stretched across multiple roles. Composes well for research-then-write, draft-then-review, plan-then-execute. Concurrent/hierarchical mode adds a manager agent (one-level hierarchy, shallower than DD-06).

Crew-scoped shared memory creates a compounding-poisoning surface with no write isolation. All agents in a crew share one memory store — a compromised agent writes a poisoned entry every other agent reads. Module 4.3 write-gating unavailable by default. Same argument as Hermes (DD-08), except compounding is across agents within a run, not across sessions. Each task-output handoff is also an untrusted-content boundary (Module 2.4 Vector 1).

After This Module

State CrewAI's defining contribution — the role-based mental model — and why 'describe the team' is the most accessible multi-agent entry point.

Distinguish emergent coordination (CrewAI) from declared coordination (LangGraph, DD-10) and explain why this contrast defines the multi-agent coordination axis (Course 4 E09).

Describe the three primitives (Crew, Agent, Task), the two orchestration modes, and the role-to-prompt mapping (Module 12 at agent granularity).

Score CrewAI 33/60 and defend the shape: 4/5 on Module 1.3, low on production-readiness modules (sandbox 1/5, verification 1/5, security 1/5).

Articulate the security implications of crew-scoped shared memory (no per-agent write isolation) and the untrusted-content boundary at each task handoff.

Artifacts

Teaching Document

Teaching document — the role-based thesis, the three primitives, emergent-vs-declared coordination axis, two orchestration modes, role-to-prompt mapping (Module 12 at agent granularity), crew-scoped memory, the security surface, the 33/60 score, anti-patterns; with learning objectives, key terms, references

READ

Diagrams

6 Mermaid/n8n diagrams — the three primitives (sequential crew with shared memory), emergent vs declared coordination axis, two orchestration modes, role-to-prompt mapping, the security surface (handoff injection + memory poisoning), the n8n sequential crew workflow

READ

Slide Deck

10 slides — reveal.js, dark theme, design-system teal; covers the describe-the-team thesis, the three primitives, emergent vs declared coordination, two orchestration modes, the security surface, the score profile, anti-patterns, the lab

READ

Teaching Script

Verbatim teaching transcript with [SLIDE N] cues, ~2,000 words spoken across 10 slide cues