The Operating Layer
Krysta Wing is a modular security fabric designed to discover, govern, and defend the integrations between autonomous AI agents and Model Context Protocol (MCP) ecosystems.
krysta_core
Vulnerability Database & Rules Engine
The foundational layer of the Krysta security ecosystem. Written in Rust, it houses the central vulnerability database and ruleset logic for inspecting Model Context Protocol (MCP) server configurations. It defines pattern-matching heuristics and CVE mappings for command injections, credential exposure, path traversals, SSRF, and description poisoning.
krysta_probe
Static Analysis Scanner CLI
The local auditing engine that scans MCP server host configurations and tool schemas for high-risk configurations. It runs static analysis checks against active servers, identifies missing authentication or potential path traversals, outputs detailed JSON summaries, and auto-generates corresponding security policy manifests.
krysta-mesh
Dynamic Security Proxy Gateway
The runtime enforcement layer that sits as a secure middleware between LLM clients and MCP servers. It intercepts tool call invocations, filters illegal argument payloads, applies granular authentication requirements, and prevents execution of unsafe commands in real-time.
krysta-npm
Developer CLI Package
The distribution CLI wrapper providing a seamless Node.js developer workflow. It enables developers to easily install, run audits, interface with the daemon, and hook security scanning directly into CI/CD pipelines and deployment operations.
Our Philosophy
As autonomous agents are given increasing permissions to read directories, access local databases, fetch external web urls, and execute arbitrary bash scripts, traditional perimeter security is no longer sufficient. Krysta Wing is built to provide an active, sandboxed, and fully auditable execution plane that ensures agent behaviors align precisely with organizational policies.