mcp_docker

The server implements a comprehensive multi-layered security framework including OAuth/OIDC authentication, IP filtering, rate limiting, audit logging (with sensitive data redaction), error sanitization, and a three-tier safety system for Docker operations (SAFE/MODERATE/DESTRUCTIVE). It includes validations for dangerous commands and volume mount paths, and automatically redacts secrets in LLM prompts. Continuous fuzzing is also utilized. However, there is a critical disconnect regarding TLS/HTTPS for the HTTP transport. While `MCP_TLS_ENABLED` configuration variables are defined, the server's HTTP transport (`uvicorn` via `FastMCP.run`) does *not* directly utilize these to provide HTTPS. The server explicitly runs on plain HTTP, with documentation recommending deployment behind a reverse proxy for TLS termination. This implementation gap means the server itself does not enforce HTTPS for HTTP connections, potentially exposing traffic in plaintext if not deployed correctly behind a proxy. This contradicts the listed `MCP_TLS_ENABLED` configuration. Additionally, container logs are returned unsanitized, posing a Retrieval Agent Deception (RADE) risk where malicious container output could be used to manipulate AI agents; users are explicitly warned to implement client-side filtering.