Skip to main content
CrewAI−7 Mixed

GILD Review

GILD's call and the buying facts behind it.

GILD analysis

CrewAI Review

PwC, Gelato and IBM Consulting describe CrewAI agents in CrewAI's own case studies, while EpicStaff removed it from its platform in September 2026. CrewAI gives a Python team role-based crews and event-driven Flows under the MIT license, on models from any vendor, and AMP hosts them free for 50 executions a month. Builders praise how fast a crew comes together and report retries, token counts, async defects and a ChromaDB dependency to settle before production.

  • Agent Frameworks
  • Agentic Development

The decision

Best fit
Python teams whose agent work maps to specialist roles, who want a managed platform with SSO and VPC deployment, on models from several vendors.
Consider another workflow
For explicit control over every step with checkpoints, evaluate LangGraph; for a small loop on OpenAI models with tracing built in, the OpenAI Agents SDK; for TypeScript web apps, the Vercel AI SDK.
Adoption condition
A pinned crewai version, idempotent tools, a telemetry decision, and token counts checked against the provider's bill before users reach the crew.

Basis: CrewAI's documentation, pricing, trust center, and repository checked Oct 28, 2026, plus 55 public accounts from builders and teams. We ran no performance benchmark. Buying facts rechecked Oct 31, 2026.

Buying & deployment factsWhat a team building on it should verify before production traffic grows.
How you buy it

Free, MIT license

AMP Basic is US$0 a month for 50 executions. Enterprise is custom-priced and adds SSO.

Source 2, CrewAI, CrewAI on GitHub: CrewAI on GitHubSource 1, CrewAI, CrewAI pricing: CrewAI pricing

Details

The crewai Python framework is free under the MIT license. CrewAI AMP Basic costs US$0 a month and includes 2 automations and 50 workflow executions per month. Enterprise is custom-priced and adds SSO with Microsoft Entra or Okta, role-based access control, and a 45-day onboarding.

Cost drivers

50 executions on Basic

AMP meters workflow executions. Enterprise bills extra executions as overage at an unpublished rate.

Source 1, CrewAI, CrewAI pricing: CrewAI pricing

Details

CrewAI AMP meters workflow executions per month. Basic stops at a maximum of 50 executions. Enterprise includes executions sized to the workflow and bills additional executions as flexible overage, with no per-execution rate published. Model tokens are a separate cost, which the AMP usage dashboard counts.

Where it runs

Python 3.10 to 3.13

AMP Basic runs on CrewAI cloud. Enterprise adds a dedicated VPC or your own infrastructure.

Source 2, CrewAI, CrewAI on GitHub: CrewAI on GitHubSource 3, CrewAI, CrewAI docs: LLMs: CrewAI docs: LLMsSource 1, CrewAI, CrewAI pricing: CrewAI pricing

Details

The framework runs on Python 3.10 to 3.13 in your own environment, and agents call models from any vendor: you set the model per agent and bring your own key for each provider. AMP Basic runs on CrewAI cloud. Enterprise deploys on CrewAI cloud, a dedicated VPC, or your own infrastructure.

What to verify

Telemetry on by default

Set OTEL_SDK_DISABLED=true to turn it off. SSO is Enterprise only.

Source 2, CrewAI, CrewAI on GitHub: CrewAI on GitHubSource 4, CrewAI, Flow HITL Management: Flow HITL ManagementSource 1, CrewAI, CrewAI pricing: CrewAI pricingSource 5, trust.crewai.com, CrewAI Trust Center: CrewAI Trust Center

Details

The library sends anonymous telemetry by default, and OTEL_SDK_DISABLED=true turns it off. AMP keeps immutable audit logs of Flow review decisions, and SSO is Enterprise only. CrewAI's Trust Center says CrewAI maintains a SOC 2 Type 2 certified security program. The Trust Center lists data retention procedures as established, without stating a retention period.

Sourced from CrewAI documentation