Mund is the security scanning component of the Weave Protocol suite, designed to vet MCP servers before you install them and detect threats in running agents. It scans for prompt injection attempts, leaked secrets, and PII exposure across tool calls and prompts. The broader Weave ecosystem includes WARD.md, a portable policy file that works across Claude Code, Google Antigravity, Microsoft Agent Framework, and MCP servers, so you write security rules once and enforce them everywhere. Mund itself focuses on the detection layer, flagging risks before they hit your agent runtime. If you're running multi-agent systems or installing third-party MCP servers, this gives you a scanning layer that catches credential leaks and injection patterns before they become incidents.
claude mcp add --transport stdio io.github.tyox-all-mund -- npx -y @weave_protocol/mund