Think of this as a security wrapper that sits between your IDE and any MCP server you're running. It intercepts every MCP tool call and response, analyzes the payloads for secrets, PII, and risky operations, then enforces policies by allowing, redacting, or blocking actions in real time. The architecture splits between local client-side redaction running on your machine and a cloud-based classifier that makes final policy decisions. You'd reach for this when using AI coding agents with access to sensitive codebases or customer data, where you need visibility and control over what gets shared with LLMs or executed via tool calls. Install it as a VS Code or Cursor extension and it automatically wraps your existing MCP setup without manual configuration.
claude mcp add --transport stdio mcpower-security-mcpower-proxy uvx mcpower-proxy