MCP Servers
Connect Guardbee security tools to Claude and Cursor
32 Model Context Protocol servers — install from npm or Smithery.
All servers
Use with Claude, Cursor, and other MCP clients
DB Gateway
@guardbee/mcp-db-gateway
KVKK/GDPR-compliant security layer between LLMs and your database — PII masking, RBAC, rate limits, and audit logs.
Security Proxy
@guardbee/mcp-security-proxy
Security gateway in front of several MCP servers — toxic-flow blocking, approvals, PII masking, and a hash-chained audit log.
Vulnerability Scanner
@guardbee/mcp-vulnerability-scanner
Start GuardBee security scans from your LLM chat, query findings, and get prioritized remediation guidance.
Compliance Checker
@guardbee/mcp-compliance-checker
Run KVKK, GDPR, and CCPA compliance checks for websites via the GuardBee API.
Secret Scanner
@guardbee/mcp-secret-scanner
Scan files and directories for exposed secrets, API keys, passwords, and other credentials.
Dependency Auditor
@guardbee/mcp-dependency-auditor
Audit npm, pip, and cargo dependencies for known CVEs against the OSV database.
SSL Inspector
@guardbee/mcp-ssl-inspector
Inspect TLS certificates, cipher suites, and protocol configurations for any domain.
DNS Intelligence
@guardbee/mcp-dns-intelligence
Enumerate DNS records, detect SPF/DMARC/DKIM misconfigurations, and find dangling subdomains.
AI Code Scanner
@guardbee/mcp-ai-code-scanner
Scan codebases for insecure LLM/AI integration patterns — client-exposed API keys, unsafe model output handling, excessive agent permissions, and PII leaks.
MCP Server Auditor
@guardbee/mcp-server-auditor
Audit other MCP servers’ tool definitions — excessive agency, dangerous sinks, loose schemas, hardcoded secrets, and wildcard CORS.
Prompt Injection Scanner
@guardbee/mcp-prompt-injection-scanner
Scan RAG content, scraped pages, and documents for indirect prompt injection — instruction overrides, role spoofing, and hidden text.
LLM Red Team
@guardbee/mcp-llm-redteam
Test guardrails on a live LLM endpoint or chatbot you own with 12 safe canary probes — jailbreaks, extraction, obfuscation, refusal suppression, and multilingual overrides.
Model Scanner
@guardbee/mcp-model-scanner
Scan ML model files (PyTorch, pickle, safetensors, Keras/H5, ONNX) for supply-chain risks — malicious pickle gadgets, disguised headers, and path traversal.
Vector Store Scanner
@guardbee/mcp-vector-store-scanner
Probe Weaviate, Qdrant, Chroma, Elasticsearch, Redis, and Postgres/pgvector endpoints for unauthenticated exposure of embeddings and RAG data.
Prompt Leak Scanner
@guardbee/mcp-prompt-leak-scanner
Catch leaked credentials and PII (API keys, national IDs, cards, IBANs) in outbound LLM prompts — on-demand MCP scans or a live reverse proxy (monitor/redact/block).
Agent Graph Auditor
@guardbee/mcp-agent-graph-auditor
Build a reachability graph across LangGraph, CrewAI, and AutoGen multi-agent configs to find transitive excessive agency via delegation or group membership.
Tool Poisoning Scanner
@guardbee/mcp-tool-poisoning-scanner
Scan other MCP servers’ tool definitions for hidden instructions in descriptions, and confused-deputy tools whose read-only name hides a shell/eval/file-write sink.
Rug Pull Detector
@guardbee/mcp-rug-pull-detector
Connect live to another MCP server, baseline tools/list, and detect an MCP rug pull — a tool’s description, schema, or annotations silently changing after approval.
Memory Poisoning Scanner
@guardbee/mcp-memory-poisoning-scanner
Scan agent code for untrusted input written into persistent cross-session memory that later returns as trusted context — MemGPT archival/core memory or a long-term vector KB.
OAuth Auditor
@guardbee/mcp-oauth-auditor
Scan an MCP server’s authorization code for OAuth 2.1 anti-patterns named in the MCP spec — token passthrough, missing audience validation, PKCE, redirect checks, and secrets.
Elicitation Auditor
@guardbee/mcp-elicitation-auditor
Scans MCP server code for elicitation anti-patterns in the 28 July 2026 spec. Flags form mode that asks for secrets or payment data, URL mode pointed at a third-party authorize endpoint, and credentials embedded in the elicitation URL.
Toxic Flow Auditor
@guardbee/mcp-toxic-flow-auditor
Audits an MCP tool catalog for toxic flows, the lethal trifecta of untrusted content, sensitive data, and exfiltration or destruction on the same server. Returns an A–F grade.
Audit Gap Auditor
@guardbee/mcp-audit-gap-auditor
Audits MCP server source for missing audit and telemetry. Flags absent tool-call trails, raw arguments in logs, a hard-disabled audit trail, silent catches, and events without correlation ids.
MCP Config Auditor
@guardbee/mcp-config-auditor
Audits Cursor, Claude, Windsurf, and VS Code MCP configs plus agent SKILL.md files. Flags unpinned packages, secrets, wildcard auto-approve, unauthenticated remotes, and shadow MCP outside an allowlist.
Threat Rules
@guardbee/mcp-threat-rules
Bridge to Agent Threat Rules. Detects prompt injection, tool poisoning, and MCP attacks at runtime, and points each match at the Guardbee scanner that should follow up.
A2A Auditor
@guardbee/mcp-a2a-auditor
Scans Agent2Agent implementations for unauthenticated push-notification webhooks, Agent Cards that require no authentication, and credentials embedded in public Agent Card metadata.
Slopsquat Scanner
@guardbee/mcp-slopsquat-scanner
Checks dependencies declared in package.json, requirements.txt, and pyproject.toml against the real npm and PyPI registries. Flags names that do not exist and names published only very recently.
Unbounded Consumption Auditor
@guardbee/mcp-unbounded-consumption-auditor
Scans LLM and agent code for unbounded consumption. Flags missing output-token limits, missing timeouts, unbounded tool-call loops, and MCP handlers with no visible rate limit.
Context Oversharing Auditor
@guardbee/mcp-context-oversharing-auditor
Audits agent and MCP source for context oversharing. Flags unrestricted session or memory dumps and context shared across users.
OWASP Scan
@guardbee/mcp-owasp-scan
One OWASP MCP Top 10 scan. Runs the Guardbee auditors against a path or a live tools/list, maps findings to MCP items, and returns an A–F grade plus SARIF.
Reverse Shield
@guardbee/mcp-reverse-shield
Audits a shipped app for what a reverse engineer gets for free. Reports exposed source maps, Electron fuses, Android debuggable and missing R8, native symbols, and missing .NET obfuscation.
Security Suite
@guardbee/security-suite
Secret scanner, dependency auditor, SSL inspector, and DNS intelligence in one MCP package.