Skip to content

future0923/debug-tools-ai

Repository files navigation

DebugTools AI

中文 | English

Validate Release npm License

DebugTools AI teaches Codex, Claude Code, OpenCode, Gemini, Cursor, Kimi, Pi, and other agents how to use DebugTools IntelliJ MCP tools to attach JVMs, generate method argument templates, invoke Java methods, and start IntelliJ run configurations with DebugTools Hotswap.

30-Second Start

Install for Codex:

curl -fsSL https://raw.githubusercontent.com/future0923/debug-tools-ai/main/install.sh | bash -s -- --codex

Verify the install:

debug-tools-ai doctor

Ask your agent:

Use DebugTools to call com.demo.UserController.getUser.

Expected agent path:

list_debug_tools_connections
generate_method_args_template   # only when parameters are needed or unclear
invoke_java_method

Hotswap start path:

list_debug_tools_run_configurations   # only when the run configuration name is unknown or ambiguous
execute_debug_tools_run_configuration

If there is no active DebugTools connection, the agent should use:

list_attachable_jvms
attach_local_jvm
invoke_java_method

What It Enables

Agents using this package can:

  • inspect current DebugTools connections
  • list attachable JVM processes
  • attach the DebugTools agent to a local JVM
  • generate DebugTools argsJson templates from Java method signatures
  • invoke Java methods through DebugTools
  • list IntelliJ run configurations for DebugTools Hotswap launch
  • start a run configuration with the DebugTools Hotswap executor
  • recover from ClassLoader issues through DebugTools HTTP when needed

Requirements

  • IntelliJ IDEA with the DebugTools plugin installed.
  • DebugTools MCP tools available from the IntelliJ plugin.
  • A target JVM that is already connected or attachable.
  • An AI client that can use the DebugTools MCP tools.

This package does not install IntelliJ IDEA, the DebugTools plugin, the DebugTools agent, or MCP server support.

Install

One-line install:

curl -fsSL https://raw.githubusercontent.com/future0923/debug-tools-ai/main/install.sh | bash -s -- --codex

Clone install:

git clone https://github.com/future0923/debug-tools-ai.git
cd debug-tools-ai
bash install.sh --codex

Install every local integration:

bash install.sh --all

After npm publication:

npm install -g debug-tools-ai
debug-tools-ai upgrade --all

Uninstall one integration:

debug-tools-ai uninstall --codex

Detailed install notes: docs/installation.md.

Supported Agents

Agent Local install
Codex bash install.sh --codex
Claude Code bash install.sh --claude
Gemini bash install.sh --gemini
OpenCode bash install.sh --opencode
Cursor bash install.sh --cursor
Kimi bash install.sh --kimi
Pi bash install.sh --pi
Generic agents Use AGENTS.md

Marketplace and host-specific installs may require host review after publication.

CLI

Command Purpose
debug-tools-ai install --codex Install one local integration.
debug-tools-ai install --all Install every supported local integration.
debug-tools-ai upgrade --all Reinstall local integrations from the current package.
debug-tools-ai uninstall --codex Remove one local integration.
debug-tools-ai uninstall --all Remove every supported local integration.
debug-tools-ai doctor Check package files and installed adapter paths.
debug-tools-ai validate Run the full validation suite.
debug-tools-ai pressure List skill behavior pressure scenarios.
debug-tools-ai pressure-report Generate a pressure run report template.
debug-tools-ai prepublish-check Run release checks before publishing.

If you are running from a checkout, use bash bin/debug-tools-ai <command>.

Doctor

Run:

debug-tools-ai doctor

OK means the expected file exists. WARN means an optional local adapter is not installed in the current HOME. FAIL means a required package file is missing, or --strict-installed was used and an installed adapter path is missing.

Strict installed-path check:

debug-tools-ai doctor --strict-installed

Usage Examples

No-arg method:

User: Use DebugTools to call com.demo.HealthController.ping.
AI: list_debug_tools_connections -> invoke_java_method

Parameterized method:

User: Call com.demo.UserController.createUser with name Codex and age 18.
AI: list_debug_tools_connections -> generate_method_args_template -> invoke_java_method

ClassLoader recovery:

ClassNotFoundException: com.demo.UserController
AI: GET /allClassLoader -> POST /classLoader/hasClass -> invoke_java_method with classLoaderIdentity

If multiple ClassLoaders can load the target class, the agent must ask the user which loader identity to use instead of guessing.

Hotswap run configuration:

User: Start the DemoApplication run configuration with DebugTools Hotswap.
AI: execute_debug_tools_run_configuration configurationName=DemoApplication

If the run configuration name is unclear, the agent should call list_debug_tools_run_configurations first. A successful execute response means startup was requested; it does not prove that DebugTools is already connected.

More examples: docs/examples.md. Install transcripts: docs/transcripts.md. Chinese docs: docs/installation-zh.md, docs/examples-zh.md, docs/transcripts-zh.md. Spring Boot demo: examples/spring-boot-demo.md.

Workflow Reference

Method invocation skill:

skills/debug-tools-method-invocation/SKILL.md

Core method invocation sequence:

list_debug_tools_connections
list_attachable_jvms
attach_local_jvm
generate_method_args_template
invoke_java_method

Hotswap skill:

skills/debug-tools-hotswap/SKILL.md

Core Hotswap sequence:

list_debug_tools_run_configurations
execute_debug_tools_run_configuration

Details:

Validation

Run:

bash scripts/validate.sh

Focused checks:

bash bin/debug-tools-ai validate
bash scripts/check-versions.sh
bash scripts/check-manifest-paths.sh
bash scripts/smoke-install.sh
bash scripts/check-pressure-scenarios.sh
bash scripts/check-release-readiness.sh

Pressure scenarios live in tests/pressure. Use them with a fresh AI agent when changing skill behavior.

Contributing and Release

Before publishing:

bash bin/debug-tools-ai prepublish-check

Release requires version updates, CHANGELOG.md, git tag, GitHub Release, and host-specific marketplace submission where applicable. The release workflow uses softprops/action-gh-release.

About

AI agent support package for DebugTools MCP workflows.

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors