java-swing-mcp

Structured AI access to running Java Swing applications. The java-swing-mcp project embeds a lightweight HTTP server inside any Swing app, exposing the live component tree, state, interactions, screenshots, and WCAG accessibility audits as JSON over localhost. One line of code turns a desktop black box into a transparent system that AI agents, test harnesses, and humans with curl can observe and control.
The core problem: legacy fat client rewrites fail because teams build against incomplete understanding of what the old application actually does. Web apps have Chrome DevTools. Desktop apps have had nothing comparable. This project closes that gap for Java Swing.
The library is open source under the MIT license. The repository includes a demo equity trading application and a Python orchestrator that connects Claude to the HTTP API for AI-driven exploration, testing, and requirements discovery.
GitHub: https://github.com/vkorost/java-swing-mcp