guardbee

Scans ML model files (PyTorch, safetensors, Keras, ONNX) for supply-chain risks

Scans MCP tool definitions for hidden instructions and confused-deputy sinks

Scans code for insecure LLM/AI integration: exposed keys, unsafe output, prompt injection

Scans Agent2Agent (A2A) protocol code for webhook SSRF, missing auth, and credential exposure

OWASP MCP08 audit/telemetry gaps: missing tool-call logs, raw args in logs, disabled audit

Scans Cursor/Claude/Windsurf/VS Code MCP configs for unpinned versions, secrets, typosquats

Scans MCP server tool definitions for excessive agency, injection sinks, hardcoded secrets

Scans MCP server auth code for OAuth 2.1 anti-patterns: token passthrough, missing audience

Catches leaked credentials and PII in outbound LLM prompts

Baselines an MCP server's tools and detects tool-definition changes after approval

Scans files for leaked secrets and API keys

Checks declared npm/PyPI dependencies against real registries to catch slopsquatting

MCP gateway: many servers, one policy, lethal-trifecta blocking, PII masking, audit log

Scans LLM/agent code for Unbounded Consumption / denial-of-wallet risks (OWASP LLM Top 10 2026 #6)

Probes vector-database endpoints for unauthenticated exposure of embeddings/RAG data

KVKK/GDPR/CCPA compliance checks for codebases

Active jailbreak/extraction/obfuscation red-teaming for live LLM endpoints

TLS certificate/cipher/protocol inspection

Triggers GuardBee scans, queries findings, AI-assisted remediation guidance

Finds transitive excessive agency across LangGraph/CrewAI/AutoGen orchestration graphs

OWASP MCP10 context oversharing: session dumps, unscoped memory, shared context, prompt leaks

CVE scanning for npm/pip/cargo dependencies via OSV