Agentshield Mcp
Scan agent inputs and flows for prompt injection, jailbreaks, and social-engineering patterns before you ship LLM features to users.
Overview
AgentShield MCP is a MCP server for the Ship phase that detects prompt injection, jailbreak, and social-engineering attacks in LLM agent traffic.
What is this MCP server?
- Detects prompt injection, jailbreak, and social-engineering attacks targeting LLM agents
- npm package @eigenart/agentshield-mcp (v0.1.3) stdio transport
- AGENTSHIELD_API_KEY from agentshield.pro signup (free tier stated)
- Purpose-built guardrail MCP, not general vulnerability scanning
- Package version 0.1.3 (@eigenart/agentshield-mcp)
- 1 required secret: AGENTSHIELD_API_KEY
- Detection focus: prompt injection, jailbreak, social-engineering (per server description)
Who is it for?
Indie builders launching customer-facing LLM agents who want a quick MCP guardrail hook in Claude Code or Cursor.
Skip if: Non-LLM apps, teams needing full pentest platforms, or environments that cannot send content to an external AgentShield API.
What do I get? / Deliverables
With AGENTSHIELD_API_KEY set, your agent can call AgentShield detection tools through @eigenart/agentshield-mcp during security review.
- stdio MCP integration with AgentShield detection API
- Agent-invokable checks for injection, jailbreak, and social-engineering patterns
Recommended MCP Servers
Journey fit
How it compares
LLM abuse-detection MCP, not a generic code review or dependency audit skill.
Common Questions / FAQ
Who is io.github.dl-eigenart/agentshield-mcp for?
Solo developers and small teams shipping agent products who need prompt-injection and jailbreak detection inside their MCP workflow.
When should I use io.github.dl-eigenart/agentshield-mcp?
Use it in the ship/security step when LLM features are about to go live or when reviewing untrusted user prompts in agent pipelines.
How do I add io.github.dl-eigenart/agentshield-mcp to my agent?
Install @eigenart/agentshield-mcp via npm, add stdio to your MCP config, and set AGENTSHIELD_API_KEY from https://agentshield.pro/signup.