Talyx

About

Your AI programming agents automatically integrate MCP servers, plugins, skills, and hooks, which are often untrusted code that can reach your shell, credentials, and source. Talyx inspects each one across 27 agents using a real tree-sitter AST and source-to-sink taint analysis, then physically prevents any malicious component from executing. The system operates entirely locally, collects no telemetry, and its source code is available.