Honeypot framework for building decoy MCP tools to detect prompt injection and malicious agent behavior.
Setup
This project documents no single launch command; follow its repository instructions.
Beelzebub
Beelzebub Labs
Related
More- MMastyfmastyf-aiOfficialSecurity proxy for LLMs and MCP that blocks prompt injection, SSRF, injection attacks and credential exfiltration.Security21 Repository stars
- KKoma Gate MCPswnotmetalOfficialClassifies input for prompt injection and out-of-scope content with the Koma Gate LLM-based classifier.Security13 Repository stars
- SShellWardjnMetaCodeChina让智能体检测提示注入、追踪敏感数据流、拦截危险命令并识别个人信息，也可检查 AI 项目的数据出境与密钥合规。Security139 Repository stars
- RReasonGatecgrtmlGateway for stdio MCP servers that blocks tool calls whose destination or content came from untrusted tool results, with an auditable reason for each decision.Security16 Repository stars
- AAletheia MCPvikasny30Filters agent tool calls before execution, blocking scope creep such as credential reads, SSRF, destructive shell or SQL commands, and prompt injection.Security12 Repository stars
- AAperion ShieldAperionLocal proxy that wraps MCP servers and blocks destructive calls such as dropping tables or recursive deletes, pins tool catalogs and scans for tool poisoning.Security11 Repository stars