injectshield MCP
v0.1.4io.github.bch1212/injectshield
Prompt-injection firewall for AI agents — scan untrusted text before LLM calls.
transport:stdio
runtime:npm
Target client
run in your project directory
claude mcp add injectshield -e INJECTSHIELD_API_KEY=<INJECTSHIELD_API_KEY> -e INJECTSHIELD_API_BASE=<INJECTSHIELD_API_BASE> -- npx -y @injectshield/mcp
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.