MCP Install Risk Scanner

MCP Install Risk Scanner highlights observable setup signals in Model Context Protocol installation commands and configuration snippets. It checks for patterns such as piped shell execution, elevated privileges, destructive commands, sensitive paths, embedded credentials, unpinned runners, insecure endpoints, broad environment access, and privileged containers. Select a command on a page and use the context menu, or open the extension and scan selected or visible setup text. Findings include redacted evidence, remediation guidance, and links to the public methodology. You can export a redacted report as JSON or Markdown for review. All analysis happens locally in the extension. There is no account, upload, telemetry, remote code, advertising, or fetched rule set. The extension requests only activeTab, scripting, and contextMenus. Page access is temporary and follows an explicit user action. A result with no flagged patterns is not proof that a package or server is safe. Static analysis cannot verify publisher identity, package contents, runtime behavior, authorization design, or obfuscated behavior. The scanner is open source under GPLv3. Its deterministic rules, versioned result schema, limitations, examples, and test fixtures are public on GitHub.