Skip to content

v0.4.0 — 8 Cybersecurity Capabilities

Choose a tag to compare

View all tags
@KryptosAI KryptosAI released this 19 Jul 02:05
· 32 commits to main since this release
v0.4.0
b8b305e

v0.4.0 — Defense-in-Depth for MCP Runtime Security

Eight new cybersecurity capabilities integrated into mcp-seatbelt:

  • RBAC: Per-agent access control with casbin (admin/agent roles, wildcard permissions)
  • Threat Intel: IOC reputation checks via ThreatFox API with 1-hour cache
  • Attack Chains: XState state machine detects multi-step attack patterns
  • Schema Validation: ajv validates tool arguments against JSON Schema, path traversal detection
  • OWASP Mapping: 13 risk rules mapped to OWASP LLM Top 10 categories
  • Compliance: SOC2, HIPAA, GDPR, ISO 27001, PCI-DSS framework tags on every policy rule
  • Honeytokens: Injects decoy credentials, detects unauthorized access
  • Forensic Capture: Records full request/response pairs as .mcpcap.json

485 tests (up from 354). 18 test files (up from 12).

📄 Read the full technical whitepaper

New CLI Commands:

  • mcp-seatbelt fuzz — Fuzz a policy against tool schemas
  • mcp-seatbelt record — Start proxy in forensic recording mode
  • mcp-seatbelt rbac-init — Initialize RBAC model and policy

Install:

npm install -g @kryptosai/mcp-seatbelt
# or
docker pull ghcr.io/kryptosai/mcp-seatbelt:latest
Assets 2
Loading