Kategori
Category
AI security
-
MCP OAuth Auditor: token passthrough & audience risks
Scan MCP server auth code for token passthrough, missing audience validation, discovery SSRF, PKCE, redirect checks, and hardcoded…
-
MCP Memory Poisoning Scanner: persistent agent memory risks
Find untrusted input written into MemGPT archival/core memory or a long-term vector KB and later recalled as trusted…
-
MCP Rug Pull Detector: catch silent tools/list drift
Connect live to an MCP server, baseline tools/list, and detect rug pulls — description/schema/annotation changes after approval —…
-
MCP Tool Poisoning Scanner: description injection & confused deputy
Find hidden instructions in MCP tool descriptions and confused-deputy tools whose read-only names hide shell/eval/file-write sinks with @guardbee/mcp-tool-poisoning-scanner.
-
MCP Agent Graph Auditor: multi-agent transitive agency
Find transitive excessive agency in LangGraph, CrewAI, and AutoGen graphs with @guardbee/mcp-agent-graph-auditor — agents that reach dangerous tools…
-
MCP Prompt Leak Scanner: stop PII/credentials in outbound prompts
Catch API keys, national IDs, cards, and IBANs in outbound LLM prompts with @guardbee/mcp-prompt-leak-scanner — MCP scans or…
-
MCP Vector Store Scanner: find open RAG databases
Probe Weaviate, Qdrant, Chroma, Elasticsearch, Redis, and Postgres/pgvector for unauthenticated exposure with @guardbee/mcp-vector-store-scanner.
-
MCP Model Scanner: ML checkpoint supply-chain scanning
Scan PyTorch, pickle, safetensors, Keras/H5, and ONNX files for malicious pickle gadgets, disguised headers, and path traversal with…
-
MCP LLM Red Team: test live chatbot guardrails
Test an LLM or chatbot you own with 12 canary probes using @guardbee/mcp-llm-redteam: jailbreak, extraction, obfuscation, and refusal…
-
GuardBee VS Code extension: see findings while coding, fix before you push
Catch secret and AI-security findings in the editor; fix them during development with Quick Fix, suppress comments, and…
-
MCP Prompt Injection Scanner: defend RAG and content
Scan RAG chunks, scraped pages, and docs for indirect prompt injection with @guardbee/mcp-prompt-injection-scanner: overrides, role spoof, hidden text,…
-
MCP Server Auditor: audit MCP tool definitions
Audit other MCP servers’ tool definitions with @guardbee/mcp-server-auditor: excessive agency, SSRF/SQL/shell sinks, loose schemas, secrets, CORS. npm +…