Java debugging tools for AI agents using JDB (Java Debugger CLI)
npx claudepluginhub brunoborges/jdb-agentic-debuggerDebug Java applications in real time using JDB (Java Debugger CLI). Attach to running JVMs or launch new ones under JDB, set breakpoints, step through code, inspect variables, analyze threads, and diagnose exceptions.
Claude Code marketplace entries for the plugin-safe Antigravity Awesome Skills library and its compatible editorial bundles.
No description available.
Curated collection of 141 specialized Claude Code subagents organized into 10 focused categories
An AI agent plugin that teaches AI agents to debug Java applications in real time using JDB — the command-line debugger shipped with every JDK. Compatible with GitHub Copilot CLI and Claude Code.
AI coding agents are already powerful analysts. They can read source code, decompile .class files, run javap to inspect bytecode, and reason about what should happen at runtime. They can attach jcmd, jstack, and jmap to a live JVM to grab thread dumps, heap histograms, and GC stats — all without JDWP. When something goes wrong, they can insert print statements, re-run the program, and parse the output.
So why give them JDB on top of all that?
Because none of those tools give an agent control over execution. They produce snapshots — what the JVM looks like right now or after the fact. But many bugs require an agent to pause the program at a precise moment and interrogate it: What is this variable's value on this line? Which branch did the code actually take? What did the object look like before the method mutated it? That's what a debugger does — and only a debugger.
Specifically, JDB gives an agent capabilities no other JDK or OS tool provides:
print order.getItems().size() or eval config.getTimeout() * 2 against real objects on the current stack frame. This isn't heap analysis — it's interactive interrogation of live state at a precise point in execution.Tools like jcmd and jstack tell an agent what the JVM is doing. JDB lets an agent control what happens next. That's the difference between monitoring and debugging — and it's why agents need both. JDB ships with every JDK, requires no IDE, and works over a simple text protocol, making it the ideal bridge between AI agents and live Java processes.
When activated, this plugin enables AI agents to:
# Add the marketplace
/plugin marketplace add brunoborges/jdb-agentic-debugger
# Install the plugin
/plugin install jdb-agentic-debugger@jdb-agentic-debugger
The .agent.md files in agents/ and the skill in skills/jdb-debugger/ are compatible with GitHub Copilot coding agent. To use them:
agents/ files into your project's .github/agents/ folderskills/jdb-debugger/ directory into your project (or reference it)The plugin includes a multi-agent chain for orchestrated Java debugging workflows. The agents are defined as .agent.md files in the agents/ directory and work with both Claude Code and GitHub Copilot.
User → JDB Debugger (orchestrator)
├── jdb-session → Interactive debugging (launch/attach, breakpoints, stepping)
├── jdb-diagnostics → Quick JVM health checks (thread dumps, deadlock detection)
└── jdb-analyst → Analysis and report consolidation