Start with the free CLI. Upgrade when your security team asks "how do we govern all our AI agents?"
| Feature | CLI | Team | Business | Enterprise |
|---|---|---|---|---|
| Agent discovery | + | + | + | + |
| Risk scoring (0-100) | + | + | + | + |
| Frameworks supported | 10 | 10 | 10 | 10 + custom |
| Agents monitored | Unlimited (local) | 25 | 100 | Unlimited |
| Web dashboard | – | + | + | + |
| Continuous monitoring | – | – | + | + |
| PDF risk reports | – | Monthly | Weekly | Real-time |
| NIST AI RMF | + | + | + | + |
| ISO/IEC 42001 | – | + | + | + |
| EU AI Act | – | – | + | + |
| OWASP Agentic Top 10 | – | – | + | + |
| CI/CD integration | – | – | + | + |
| Auditor-ready export | – | – | + | + |
| SSO / SAML | – | – | – | + |
| Insurance data API | – | – | – | + |
| Support | Community | Priority | Dedicated CSM |
Snyk agent-scan is a runtime security scanner — it detects vulnerabilities in agent code. Agent Shield is an audit and compliance tool — it discovers agents, scores their risk posture, maps findings to compliance frameworks (NIST, ISO 42001, EU AI Act), and generates board-ready reports. They're complementary, not competitive. Think of Snyk as your firewall and Agent Shield as your auditor.
LangChain, LangGraph, CrewAI, AutoGen (Microsoft), OpenAI Assistants, Anthropic MCP, Anthropic Agent SDK, AWS Bedrock Agents, Vercel AI SDK, and custom agents with standard patterns (system prompts, tool calls, agent loops). We add new frameworks regularly based on community requests.
You're charged per unique AI agent monitored per month. An "agent" is a distinct AI system with its own tools, permissions, and system prompt. A single codebase might have 3 agents or 300 — you only pay for what's there. The CLI is always free for local scans.
Yes. Business and Enterprise tiers map every finding to EU AI Act articles (Article 9 Risk Management, Article 10 Data Governance, Article 12 Record-keeping, Article 14 Human Oversight). The auditor-ready export gives your compliance team exactly what they need to demonstrate conformity for high-risk AI systems. High-risk obligations take effect August 2026.
Yes. Agent Shield CLI is open source under the MIT license. No account required. No usage limits. No phone-home telemetry. The paid tiers add a web dashboard, continuous monitoring, team collaboration, compliance framework mapping beyond NIST, and auditor-ready reporting. The CLI is the foundation — it will never be paywalled.
Enterprise customers can export structured risk assessment data via API for use in AI insurance underwriting. As the AI insurance market develops (projected $4.7B-$500B by 2030), this data becomes the actuarial foundation that insurers need to price AI liability policies. Early customers build the dataset that makes AI insurance possible.