MCPShield Agent

MCPShield is an AI agent security platform for finding and assessing MCP servers across an organization. It is designed around a stated visibility problem: security teams cannot easily see which MCP servers are running, what AI tools can reach them, or which configurations expose databases, file systems, APIs, credentials, and other sensitive systems.

The product scans for MCP configurations and automatically discovers MCP servers on endpoints. It analyzes each server’s configuration and assigns a risk score from 0 to 100, with severity labels such as Critical, High, Medium, and Low. Supported checks mentioned on the site include database access patterns, sensitive environment variables, and file system scope permissions. It also offers on-demand scanning, uploads results to a dashboard, and shows live activity timelines and instant alerts for high-risk configurations. The dashboard includes detailed breakdowns of risk factors, and the site also describes dynamic risk score badges for internal documentation. A GitHub Action is shown for CI use, where pull requests that touch MCP configs can be risk-scored automatically.

MCPShield’s discovery flow is described for Claude Desktop, Cursor, VS Code, and custom setups, and the site says it detects Claude, ChatGPT, Cursor, and custom MCP. It is presented as cross-platform for Windows, macOS, and Linux. The product can be installed as a lightweight Python agent, with a pip command shown for installation, and the scan flow includes configuring the agent with an API key before running a scan. The site also advertises a live demo and says some scanning can run entirely in the browser without signup.

The page says a free tier is available, and the agent is MIT licensed. It also describes the product as privacy-first and says it never captures credential values.