AgentShield
Rank #7797glama/elliotllliu/agent-shield
Full-stack security for AI agents — static analysis + MCP runtime interception. 31 rules detect prompt injection, data exfiltration, backdoors, tool poisoning, and cross-file attack chains. Includes MCP proxy for real-time blocking, Python AST taint tracking, multi-language injection detection (8 languages), and AI-powered deep analysis. Free, offline, zero-config.
AgentShield is a Model Context Protocol (MCP) server published by elliotllliu. It ranks #7797 of 58,832 servers tracked on MCP Toplist, and its repository has 13 GitHub stars. AgentShield is listed on Glama, and ships as a single rolling release with no explicit version metadata. It was first listed on Mar 13, 2026.
Ranks ahead of 51,035 of 58,832 servers on MCP Toplist.
Listed on 1 registry
elliotllliu
Not versioned
This server is published through a registry that does not expose explicit version metadata. The listing tracks a single rolling release.
Frequently asked questions
- What is AgentShield?
- Full-stack security for AI agents — static analysis + MCP runtime interception. 31 rules detect prompt injection, data exfiltration, backdoors, tool poisoning, and cross-file attack chains. Includes MCP proxy for real-time blocking, Python AST taint tracking, multi-language injection detection (8 languages), and AI-powered deep analysis. Free, offline, zero-config.
- Who maintains AgentShield?
- AgentShield is maintained by elliotllliu, which publishes 1 MCP server (0 total versions) tracked on MCP Toplist.
- Is AgentShield an official MCP server?
- AgentShield is not on the Official MCP Registry. It is listed on Glama.
- How many versions does AgentShield have?
- AgentShield ships as a single rolling release with no explicit version metadata.
- Where can I find the source code for AgentShield?
- The source code for AgentShield is hosted at github.com/elliotllliu/agent-shield.