SuperAPI AI Directory

jetbrains-debugger-mcp-plugin

JetBrains IDE plugin giving coding agents programmatic control of the debugger.

Open website

About this tool

Purpose: This is a JetBrains IDE plugin that exposes an MCP (Model Context Protocol) server within the IDE, allowing AI coding assistants to fully control the debugger programmatically and achieve autonomous debugging.

Main capabilities: Debug session management (start/stop run configurations, retrieve variables/stack/source context in a single call, concurrent multiple sessions); breakpoint management (line breakpoints, conditional breakpoints, Tracepoints that do not pause); execution control (Step Over/Into/Out, resume and pause, run to a specified line, jump to a specified line); variable inspection (view local variables, parameters, object fields, and modify variable values during debugging); expression evaluation (execute arbitrary expressions and multi-line code snippets in the current context, with safety controls); stack and thread navigation (full call stack, switch stack frames, list thread states). A total of 24 MCP tools are provided, along with a companion skill that guides AI agents in using the debugger tools, which can be installed to .claude/skills/ with one click or exported as .skill / .zip.

Usage: In the IDE, install via Settings/Preferences > Plugins > Marketplace by searching for "Debugger MCP Server", or install from the JetBrains Marketplace, or manually install from disk and restart the IDE. After opening a project, the MCP server automatically starts on an IDE-specific port, which can be viewed in Settings > Tools > Debugger MCP Server (for example, 29190 for IntelliJ IDEA). Configure the AI assistant with the server URL:http://127.0.0.1:{PORT}/debugger-mcp/streamable-http。也可在底部 In the "Debugger MCP Server" tool window, click the "Install on Coding Agents" button to automatically install for Claude Code CLI and Codex CLI, or copy the JSON configuration for clients such as Gemini CLI and Cursor. When multiple projects are open, use the project_path parameter to specify the project.

Use cases: Suitable for AI-assisted development workflows where you want the AI assistant to autonomously investigate defects, verify fixes, or explore code behavior.

Limitations: Requires JetBrains IDE 2025.2 or later, and JVM 21 or later. CI validation supports IntelliJ IDEA, PyCharm, WebStorm, and GoLand; community testing supports RustRover, Android Studio, and PhpStorm; RubyMine, CLion, DataGrip, etc. may work but are untested. The safety settings of evaluate_expression are best-effort guardrails rather than a sandbox; in Unrestricted mode, expressions can call methods, modify state, and access files/network.