Gatekeeper
Provides secure and controlled execution of shell commands and HTTP proxy capabilities for AI assistants.
Gatekeeper is a robust MCP (Model Context Protocol) server designed to enhance the security and control over shell command execution and HTTP proxying for AI assistants. It offers configurable plugin definitions, multiple sandboxing options (none, bubblewrap, WASM), and policy-based argument validation, ensuring that AI agents can interact with system resources safely and predictably. With support for API key authentication, rate limiting, audit logging, and rich UI rendering, it empowers developers to integrate powerful tooling while maintaining strict oversight and a user-friendly experience.
주요 기능
01Policy-based access control using glob patterns for argument validation
02Supports stdio, HTTP (JSON-RPC 2.0 with Bearer token auth), and bridge operating modes
03Optional SQLite-based audit logging for all tool calls and MCP Apps UI support
040 GitHub stars
05Flexible JSON-based plugin configuration for defining tools and allowed environments
06Secure command execution with multiple sandboxing modes (none, bubblewrap, WASM)
사용 사례
01Integrating secure command execution directly with MCP clients like Claude Desktop
02Exposing controlled shell commands as a rate-limited and authenticated HTTP API
03Proxying existing stdio MCP servers over HTTP, adding security and audit capabilities