MCP JS Debugger
Enables AI assistants to debug JavaScript and TypeScript applications by connecting to Chrome DevTools Protocol-compatible debuggers, allowing them to set breakpoints, step through code, inspect variables, and evaluate expressions with full source map support.
README
MCP JS Debugger
An MCP (Model Context Protocol) server that exposes Chrome DevTools Protocol debugging capabilities, enabling AI assistants to debug JavaScript and TypeScript applications.
Features
- Connect to any CDP-compatible debugger (Node.js, Chrome, Edge)
- Set, list, and remove breakpoints
- Step through code (over, into, out)
- Inspect call stacks with source map support
- Evaluate expressions in any stack frame
- View and modify variables
- Pause on exceptions
- Full source map support for debugging transpiled code
Installation
npm install
npm run build
Claude Code integration
Add to your Claude Code configuration:
claude mcp add mcp-js-debugger -- npx mcp-js-debugger
Or add to .mcp.json:
{
"mcpServers": {
"mcp-js-debugger": {
"command": "npx",
"args": ["mcp-js-debugger"]
}
}
}
Usage
Starting a debug target
Start your Node.js application with the inspector:
# Pause on first line (recommended for setting initial breakpoints)
node --inspect-brk=9229 your-script.js
# Or start without pausing
node --inspect=9229 your-script.js
Available tools
| Tool | Description |
|---|---|
connect_debugger |
Connect to a CDP endpoint via WebSocket URL |
disconnect_debugger |
Disconnect from a debugging session |
set_breakpoint |
Set a breakpoint by URL and line number |
remove_breakpoint |
Remove a breakpoint by ID |
list_breakpoints |
List all breakpoints in a session |
resume_execution |
Resume execution after pause |
step_over |
Step over the current statement |
step_into |
Step into a function call |
step_out |
Step out of the current function |
pause_execution |
Pause running execution |
get_call_stack |
Get the current call stack with source locations |
evaluate_expression |
Evaluate a JavaScript expression |
get_scope_variables |
Get variables in a scope |
set_variable_value |
Modify a variable's value |
set_pause_on_exceptions |
Configure exception handling |
get_original_location |
Map generated to original source location |
get_script_source |
Get script source code |
list_scripts |
List loaded scripts |
Example workflow
-
Start your application with
--inspect-brk:node --inspect-brk=9229 app.js -
Get the WebSocket URL:
curl http://localhost:9229/json -
Connect the debugger:
connect_debugger(websocket_url: "ws://localhost:9229/<id>") -
Set breakpoints:
set_breakpoint(session_id: "...", url: "file:///path/to/app.js", line_number: 10) -
Resume execution to hit the breakpoint:
resume_execution(session_id: "...") -
Inspect state when paused:
get_call_stack(session_id: "...") get_scope_variables(session_id: "...", call_frame_id: "...", scope_index: 0) evaluate_expression(session_id: "...", expression: "myVariable") -
Continue debugging:
step_over(session_id: "...") resume_execution(session_id: "...")
Source map support
The server automatically loads source maps for transpiled code (TypeScript, bundled JavaScript, etc.). When source maps are available:
- Call stacks show original source locations
- Breakpoints can be set on original source files
get_original_locationmaps generated positions to original sourceget_script_sourcecan return original source content
Development
# Build
npm run build
# Run tests
npm test
# Run tests with coverage
npm run test:coverage
# Watch mode for development
npm run dev
Architecture
- cdp-client.ts - Low-level Chrome DevTools Protocol client wrapper
- session-manager.ts - Manages multiple debugging sessions
- source-map-manager.ts - Handles source map loading and position mapping
- server.ts - MCP server implementation with tool handlers
- types.ts - TypeScript type definitions
Requirements
- Node.js 18.0.0 or later
- A CDP-compatible debug target (Node.js, Chrome, Edge, etc.)
Licence
Apache-2.0
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。