Skip to content
mcp/harbor
Verified official

agent immune MCP

v0.2.2

io.github.denial-web/agent-immune

AI agent security: prompt injection detection, semantic memory, output scanning, prompt hardening

transport:stdio runtime:pypi
View source

Target client

run in your project directory 

claude mcp add agent-immune -- uvx agent-immune

Adds it for this project only. Append --scope user to make it available everywhere. Claude Code docs

This listing does not declare its tools. Connect the server and your client will discover them on the handshake.