DevTools Debugger MCP
Exposes full Chrome DevTools Protocol debugging capabilities, including breakpoints, call stacks, and source maps.
Node.js Debugger MCP
An MCP server that provides comprehensive Node.js debugging capabilities using the Chrome DevTools Protocol. This server enables AI assistants to debug Node.js applications with full access to breakpoints, stepping, variable inspection, call stacks, expression evaluation, and source maps.
Why use this MCP server?
This MCP server is useful when you need AI assistance with debugging Node.js applications. It provides programmatic access to all the debugging features you'd find in Chrome DevTools or VS Code, allowing AI assistants to help you set breakpoints, inspect variables, step through code, and analyze runtime behavior.
Features
- Full Node.js debugger: Set breakpoints, conditional breakpoints, logpoints, and pause-on-exceptions
- Stepping controls: Step over/into/out, continue to location, restart frame
- Variable inspection: Explore locals/closure scopes,
thispreview, and drill down into object properties - Expression evaluation: Evaluate JavaScript expressions in the current call frame with console output capture
- Call stack analysis: Inspect call stacks and pause-state information
- Source map support: Debug TypeScript and other transpiled code with full source map support
- Console monitoring: Capture and review console output during debugging sessions
Installation
npm install devtools-debugger-mcp
Configuration
Add the server to your MCP settings configuration:
{
"devtools-debugger-mcp": {
"command": "node",
"args": ["path/to/devtools-debugger-mcp/dist/index.js"]
}
}
Alternatively, if installed globally, you can use the CLI binary:
{
"devtools-debugger-mcp": {
"command": "devtools-debugger-mcp"
}
}
Node.js Debugging
This MCP server can debug Node.js programs by launching your script with the built‑in inspector (--inspect-brk=0) and speaking the Chrome DevTools Protocol (CDP).
How it works
start_node_debugspawnsnode --inspect-brk=0 your-script.js, waits for the inspector WebSocket, attaches, and returns the initial pause (first line) with apauseIdand top call frame.- You can then set breakpoints (by file path or URL regex), choose pause-on-exceptions, and resume/step. At each pause, tools can inspect scopes, evaluate expressions, and read console output captured since the last step/resume.
- When the process exits, the server cleans up the CDP session and resets its state.
Quickstart (from an MCP-enabled client)
- Start a debug session
{ "tool": "start_node_debug", "params": { "scriptPath": "/absolute/path/to/app.js" } }
- Set a breakpoint (file path + 1-based line)
{ "tool": "set_breakpoint", "params": { "filePath": "/absolute/path/to/app.js", "line": 42 } }
- Run to next pause (optionally include console/stack)
{ "tool": "resume_execution", "params": { "includeConsole": true, "includeStack": true } }
- Inspect at a pause
{ "tool": "inspect_scopes", "params": { "maxProps": 15 } }
{ "tool": "evaluate_expression", "params": { "expr": "user.name" } }
- Step
{ "tool": "step_over" }
{ "tool": "step_into" }
{ "tool": "step_out" }
- Finish
{ "tool": "stop_debug_session" }
Node.js tool reference (summary)
start_node_debug({ scriptPath, format? })— Launches Node with inspector and returns initial pause.set_breakpoint({ filePath, line })— Breakpoint by file path (1-based line).set_breakpoint_condition({ filePath?, urlRegex?, line, column?, condition, format? })— Conditional breakpoint or by URL regex.add_logpoint({ filePath?, urlRegex?, line, column?, message, format? })— Logpoint via conditional breakpoint that logs and returnsfalse.set_exception_breakpoints({ state })—none | uncaught | all.blackbox_scripts({ patterns })— Ignore frames from matching script URLs.list_scripts()/get_script_source({ scriptId? | url? })— Discover and fetch script sources.continue_to_location({ filePath, line, column? })— Run until a specific source location.restart_frame({ frameIndex, pauseId?, format? })— Re-run the selected frame.resume_execution({ includeScopes?, includeStack?, includeConsole?, format? })— Continue to next pause or exit.step_over|step_into|step_out({ includeScopes?, includeStack?, includeConsole?, format? })— Stepping with optional context in the result.evaluate_expression({ expr, pauseId?, frameIndex?, returnByValue?, format? })— Evaluate in a paused frame; defaults to top frame.inspect_scopes({ maxProps?, pauseId?, frameIndex?, includeThisPreview?, format? })— Locals/closures andthissummary.get_object_properties({ objectId, maxProps?, format? })— Drill into object previews.list_call_stack({ depth?, pauseId?, includeThis?, format? })— Top N frames summary.get_pause_info({ pauseId?, format? })— Pause reason/location summary.read_console({ format? })— Console messages since the last step/resume.stop_debug_session()— Kill process and detach.
Notes
- File paths are converted to
file://URLs internally for CDP compatibility. lineis 1-based; CDP is 0-based internally.- The server buffers console output between pauses; fetch via
includeConsoleon step/resume or withread_console. - Use
set_output_format({ format: 'text' | 'json' | 'both' })to set default response formatting.
Available Tools
This MCP server provides the following Node.js debugging tools. All tools support optional format parameter ('text' or 'json') to control response formatting.
Session Management
start_node_debug- Launch a Node.js script with debugging enabledstop_debug_session- Terminate the debugging session and clean up
Breakpoint Management
set_breakpoint- Set a breakpoint at a specific file and lineset_breakpoint_condition- Set a conditional breakpoint or breakpoint by URL regexadd_logpoint- Add a logpoint that logs messages when hitset_exception_breakpoints- Configure pause-on-exception behavior
Execution Control
resume_execution- Continue execution to the next breakpoint or completionstep_over- Step over the current linestep_into- Step into function callsstep_out- Step out of the current functioncontinue_to_location- Run until reaching a specific locationrestart_frame- Restart execution from a specific call frame
Inspection and Analysis
inspect_scopes- Examine local variables, closures, andthiscontextevaluate_expression- Evaluate JavaScript expressions in the current contextget_object_properties- Drill down into object propertieslist_call_stack- View the current call stackget_pause_info- Get information about the current pause state
Utilities
list_scripts- List all loaded scriptsget_script_source- Retrieve source code for scriptsblackbox_scripts- Configure scripts to skip during debuggingread_console- Read console output captured during debugging
For detailed usage examples and parameter descriptions, see the "Node.js Debugging" section above.
License
MIT
Related Servers
Scout Monitoring MCP
sponsorPut performance and error data directly in the hands of your AI assistant.
Alpha Vantage MCP Server
sponsorAccess financial market data: realtime & historical stock, ETF, options, forex, crypto, commodities, fundamentals, technical indicators, & more
Claude Swarm MCP Server
An MCP server for multi-agent orchestration using Claude AI via Claude Desktop.
Reference Servers
Reference implementations of Model Context Protocol (MCP) servers in Typescript and Python, showcasing MCP features and SDK usage.
Claude Code MCP
Orchestrates multiple Claude Code agents across iTerm2 sessions, providing centralized management and inter-agent communication.
Pica MCP Server
An MCP server for Pica that enables seamless interaction with various third-party services through a standardized interface.
OpenZeppelin MCP
Access secure, standards-compliant smart contract templates from OpenZeppelin, including ERC20, ERC721, and ERC1155.
Projet MCP Server-Client
An implementation of the Model Context Protocol (MCP) for communication between AI models and external tools, featuring server and client examples in Python and Spring Boot.
Markdown Sidecar MCP Server
Serve and access markdown documentation for locally installed NPM, Go, or PyPi packages.
ndlovu-code-reviewer
Manual code reviews are time-consuming and often miss the opportunity to combine static analysis with contextual, human-friendly feedback. This project was created to experiment with MCP tooling that gives AI assistants access to a purpose-built reviewer. Uses the Gemini cli application to process the reviews at this time and linting only for typescript/javascript apps at the moment. Will add API based calls to LLM's in the future and expand linting abilities. It's also cheaper than using coderabbit ;)
Fastn Server
A scalable platform for dynamic tool registration and execution based on API definitions, with integrations for services like Claude.ai and Cursor.ai.
Gemini MCP
An MCP server that orchestrates Google Gemini and Claude Code models via the OpenRouter API.