Loading the catalogue…
Agent frameworks · CrewAI Inc. · US
CrewAI
US CLOUD Act exposureOpen sourceMIT
STAV REVIEW · LAST REVIEWED 06/21/2026
“Genuinely Sovereign in the self-hosted path, but five critical CVEs in twelve months and a missing Enterprise DPA make the caution rating earned, not precautionary.”
Good for
- Platform engineering teams building self-hosted multi-agent pipelines who need direct Inference routing to their own LLM provider with no vendor intermediation
- Organisations requiring on-premises or air-gapped agent orchestration under a permissive MIT licence with no copyleft obligations
- Teams that want an actively maintained framework — roughly one release every 1–2 days, 100+ commits from 25 contributors in 90 days — and can pin versions for security compliance
- Enterprises already evaluating SSO and audit-log requirements, where the framework's documented Enterprise-tier features provide a starting point
Watch out for
- The Enterprise cloud platform (app.crewai.com) has no publicly accessible DPA, no documented inference data-retention terms, and no confirmed sub-processor list — teams subject to GDPR or sector-specific regulation must not route workloads through it until a signed agreement is in place
- Five CVE-tagged events in twelve months including a CVSS 9.2 supply-chain exposure and four critical CodeInterpreterTool CVEs (RCE, file read, SSRF, Docker escape via CERT/CC VU#221883); all confirmed remediated at v1.14.5+, but the density warrants pinning and ongoing monitoring
- The OpenTelemetry SDK is a hard non-optional dependency — anonymous telemetry ships on by default and requires an explicit OTEL_SDK_DISABLED=true opt-out; setting share_crew=True would additionally send prompt content to CrewAI
Editorial assessment, not legal advice. Stav's risk ratings, scores, and verdicts are our own analysis of publicly available information and may be incomplete or out of date. Verify independently before making compliance or procurement decisions.
MODEL REQUIREMENTS
What CrewAI needs from a model.
Required
- Tool calling
Recommended
- Extended thinking
- Long context (≥128k)
- Structured output
Known issues8 open
Public security, privacy and governance advisories that are still open. Each row carries a source link — these are factual claims, not editorial judgements.
2026-06-30
lowUpstream change on crewai: github_release '1.14.7' → '1.15.1', github_last_push '2026-06-23T05:31:42Z' → '2026-06-30T05:31:53Z'
other · open
Source ↗2026-06-23
lowUpstream change on crewai: github_last_push '2026-06-16T05:00:21Z' → '2026-06-23T05:31:42Z'
other · open
Source ↗2026-06-16
lowUpstream change on crewai: github_release '1.14.6' → '1.14.7', github_last_push '2026-06-09T05:04:18Z' → '2026-06-16T05:00:21Z'
other · open
Source ↗2026-06-09
lowUpstream change on crewai: github_last_push '2026-06-02T05:41:53Z' → '2026-06-09T05:04:18Z'
other · open
Source ↗2026-06-02
lowUpstream change on crewai: github_release '1.14.5' → '1.14.6', github_last_push '2026-05-18T23:47:26Z' → '2026-06-02T05:41:53Z'
other · open