Node.js debugging in VS Code

The Visual Studio Code editor has built-in debugging support for the Node.js runtime and can debug JavaScript, TypeScript, and many other languages that are transpiled into JavaScript. Setting up a project for Node.js debugging is usually straightforward with VS Code providing appropriate launch configuration defaults and snippets.

Note: If you are just getting started with VS Code, you can learn about general debugging features and creating launch.json configuration files in the Debugging topic.

This Node.js debugging document goes into more detail about configurations and features for more advanced debugging scenarios. You'll find instruction for debugging with source maps, stepping over external code, doing remote debugging, and much more.

If you'd like to watch an introductory video, see Getting started with Node.js debugging.

Launch configuration attributes

Debugging configurations are stored in a launch.json file located in your workspace's .vscode folder. An introduction into the creation and use of debugging configuration files is in the general Debugging topic.

Below is a reference on launch.json attributes specific to the Node.js debugger.

The following attributes are supported in launch configurations of type launch and attach:

These attributes are only available for launch configurations of request type launch:

This attribute is only available for launch configurations of request type attach:

Launch configurations for common scenarios

You can trigger IntelliSense (kb(editor.action.triggerSuggest)) in your launch.json file to see launch configuration snippets for commonly used Node.js debugging scenarios.

Launch configuration snippets for Node.js

You can also bring up the snippets with the Add Configuration... button in the lower right of the launch.json editor window.

Add Configuration button

Here is the list of all snippets:

Node console

By default, Node.js debug sessions launch the target in the internal VS Code Debug Console. Since the Debug Console does not support programs that need to read input from the console, you can enable either an external terminal or use the VS Code Integrated Terminal by setting the console attribute in your launch configuration to externalTerminal or integratedTerminal respectively. The default is internalConsole.

If an external terminal is used, you can configure which terminal program to use via the terminal.external.windowsExec, terminal.external.osxExec, and terminal.external.linuxExec settings.

Launch configuration support for 'npm' and other tools

Instead of launching the Node.js program directly with node, you can use 'npm' scripts or other task runner tools directly from a launch configuration:

Let's look at an 'npm' example. If your package.json has a 'debug' script, for example:

  "scripts": {
    "debug": "node --nolazy --inspect-brk=9229 myProgram.js"
  },

the corresponding launch configuration would look like this:

{
    "name": "Launch via npm",
    "type": "node",
    "request": "launch",
    "cwd": "${workspaceFolder}",
    "runtimeExecutable": "npm",
    "runtimeArgs": [
        "run-script", "debug"
    ],
    "port": 9229
}

Multi version support

If you are using 'nvm' (or 'nvm-windows') to manage your Node.js versions, it is possible to specify a runtimeVersion attribute in a launch configuration for selecting a specific version of Node.js:

{
    "type": "node",
    "request": "launch",
    "name": "Launch test",
    "runtimeVersion": "7.10.1",
    "program": "${workspaceFolder}/test.js"
}

If you are using 'nvs' to manage your Node.js versions, it is possible to use runtimeVersion attribute to select a specific version of Node.js:

{
    "type": "node",
    "request": "launch",
    "name": "Launch test",
    "runtimeVersion": "chackracore/8.9.4/x64",
    "program": "${workspaceFolder}/test.js"
}

Make sure to have those Node.js versions installed that you want to use with the runtimeVersion attribute as the feature will not download and install the version itself. So you will have to run something like nvm install 7.10.1 or nvs add 7.10.1 from the integrated terminal if you plan to add "runtimeVersion": "7.10.1" to your launch configuration.

Note: If VS Code detects that "nvs" is installed, it does not fall back to "nvm" if a specific Node.js version cannot be found in "nvs". Using both "nvs" and "nvm" at the same time is not supported.

Load environment variables from external file (node)

The VS Code Node debugger supports loading environment variables from a file and passing them to the Node.js runtime. To use this feature, add an attribute envFile to your launch configuration and specify the absolute path to the file containing the environment variables:

   //...
   "envFile": "${workspaceFolder}/.env",
   "env": { "USER": "john doe" }
   //...

Any environment variable specified in the env dictionary will override variables loaded from the file.

Here is an example of an .env file:

USER=doe
PASSWORD=abc123

# a comment

# an empty value:
empty=

# new lines expanded in quoted strings:
lines="foo\nbar"

Attaching to Node.js

If you want to attach the VS Code debugger to a Node.js program, launch Node.js in VS Code's integrated terminal as follows:

node --inspect program.js

or if the program should not start running but must wait for the debugger to attach:

node --inspect-brk program.js

Now you have three options for attaching the debugger to your program:

Let's go through these options in detail:

Auto Attach Feature

If the Auto Attach feature is enabled, the Node debugger automatically attaches to Node.js processes that have been launched in debug mode from VS Code's Integrated Terminal.

To enable the feature, either use the Toggle Auto Attach action or, if the Node debugger is already activated, use the Auto Attach Status Bar item.

After enabling Auto Attach the debugger should attach to your program within a second:

Auto Attach

Whether or not a process is in "debug mode" is determined by analyzing the program arguments. Currently, we detect the patterns --inspect, --inspect-brk, --inspect-port, --debug, --debug-brk, --debug-port (all optionally followed by a '=' and a port number).

Note: this feature does not (yet) work for terminal multiplexers like tmux (where launched processes are not children of VS Code's integrated terminal).

Attach to Node Process action

The Attach to Node Process action opens a Quick Pick menu that lists all potential processes that are available to the Node.js debugger:

Node.js Process picker

The individual processes listed in the picker show the debug port and the detected protocol in addition to the process ID. You should find your Node.js process in that list and after selecting it, the Node.js debugger tries to attach to it.

In addition to Node.js processes, the picker also shows other programs that were launched with one of the various forms of --debug or --inspect arguments. This makes it possible to attach to Electron's or VS Code's helper processes.

Setting up an "Attach" configuration

This option requires more work but in contrast to the previous two options it allows you to configure various debug configuration options explicitly.

The simplest "attach" configuration looks like this:

{
    "name": "Attach to Process",
    "type": "node",
    "request": "attach",
    "port": 9229
}

The port 9229 is the default debug port of the --inspect and --inspect-brk options. To use a different port (for example 12345), add it to the options like this: --inspect=12345 and --inspect-brk=12345 and change the port attribute in the launch configuration accordingly.

If you want to attach to a Node.js process that hasn't been started in debug mode, you can do this by specifying the process ID of the Node.js process as a string:

{
    "name": "Attach to Process",
    "type": "node",
    "request": "attach",
    "processId": "53426"
}

Since it is a bit laborious to repeatedly find the process ID and enter it in the launch configuration, Node debug supports a command variable PickProcess that binds to the process picker (from above) and that lets you conveniently pick the process from a list of Node.js processes.

By using the PickProcess variable the launch configuration looks like this:

{
    "name": "Attach to Process",
    "type": "node",
    "request": "attach",
    "processId": "${command:PickProcess}"
}

Stop debugging

Using the Debug: Stop action (available in the Debug toolbar or via the Command Palette) stops the debug session.

If the debug session was started in "attach" mode (and the red terminate button in the Debug toolbar shows a superimposed "plug"), pressing Stop disconnects the Node.js debugger from the debuggee that then continues execution.

If the debug session is in "launch" mode, pressing Stop does the following:

  1. When pressing Stop for the first time, the debuggee is requested to shutdown gracefully by sending a SIGINT signal. The debuggee is free to intercept this signal and clean up anything as necessary and then shut down. If there are no breakpoints (or problems) in that shutdown code, the debuggee and the debug session will terminate.

  2. However if the debugger hits a breakpoint in the shutdown code or if the debuggee does not terminate properly by itself, then the debug session will not end. In this case, pressing Stop again will force terminate the debuggee and its child processes (SIGKILL).

So if you see that a debug session doesn't end when you press the red Stop button, then press the button again to force a shutdown of the debuggee.

Note that on the Windows operating system, pressing Stop always forcibly kills the debuggee and its child processes.

Source maps

The Node.js debugger of VS Code supports JavaScript source maps that help debugging of transpiled languages, for example, TypeScript or minified/uglified JavaScript. With source maps, it is possible to single step through or set breakpoints in the original source. If no source map exists for the original source or if the source map is broken and cannot successfully map between the source and the generated JavaScript, then breakpoints show up as unverified (gray hollow circles).

Source maps can be generated with two kinds of inlining:

VS Code supports both the inlined source maps and the inlined source.

The source map feature is controlled by the sourceMaps attribute that defaults to true. This means that node debugging always tries to use source maps (if it can find any) and as a consequence, you can even specify a source file (for example, app.ts) with the program attribute.

If you need to disable source maps for some reason, you can set the sourceMaps attribute to false.

If the generated (transpiled) JavaScript files do not live next to their source but in a separate directory, you must help the VS Code debugger locating them by setting the outFiles attribute. This attribute takes multiple glob patterns for including and excluding files from the set of generated JavaScript files. Whenever you set a breakpoint in the original source, VS Code tries to find the generated JavaScript code in the files specified by outFiles.

Since source maps are not automatically created, you must configure the transpiler you are using to create them. For TypeScript this can be done in the following way:

tsc --sourceMap --outDir bin app.ts

This is the corresponding launch configuration for a TypeScript program:

{
    "version": "0.2.0",
    "configurations": [
        {
            "name": "Launch TypeScript",
            "type": "node",
            "request": "launch",
            "program": "app.ts",
            "outFiles": [ "${workspaceFolder}/bin/**/*.js" ]
        }
    ]
}

Smart stepping

With the smartStep attribute set to true in a launch configuration, VS Code will automatically skip 'uninteresting code' when stepping through code in the debugger. 'Uninteresting code' is code that is generated by a transpiling process but is not covered by a source map so it does not map back to the original source. This code gets in your way when stepping through source code in the debugger because it makes the debugger switch between the original source code and generated code that you are not really interested in. smartStep will automatically step through code not covered by a source map until it reaches a location that is covered by a source map again.

This is especially useful for cases like async/await downcompilation in TypeScript, where the compiler injects helper code that is not covered by a source map.

Please note that the smartStep feature only applies to JavaScript code that was generated from source and therefore has a source map. For JavaScript without sources, the smart stepping option has no effect.

JavaScript source map tips

A common issue when debugging with source maps is that you'll set a breakpoint, and it will turn gray. If you hover the cursor over it, you'll see the message, "Breakpoint ignored because generated code not found (source map problem?)". What now? There are a range of issues that can lead to this. First, a quick explanation of how the Node debug adapter handles source maps.

When you set a breakpoint in app.ts, the debug adapter has to figure out the path to app.js, the transpiled version of your TypeScript file, which is what is actually running in Node. But, there is not a straightforward way to figure this out starting from the .ts file. Instead, the debug adapter uses the outFiles attribute in the launch.json to find all the transpiled .js files, and parses them for a source map, which contains the locations of its associated .ts files.

When you build your app.ts file in TypeScript with source maps enabled, it either produces an app.js.map file, or a source map inlined as a base64-encoded string in a comment at the bottom of the app.js file. To find the .ts files associated with this map, the debug adapter looks at two properties in the source map, sources, and sourceRoot. sourceRoot is optional - if present, it is prepended to each path in sources, which is an array of paths. The result is an array of absolute or relative paths to .ts files. Relative paths are resolved relative to the source map.

Finally, the debug adapter searches for the full path of app.ts in this resulting list of .ts files. If there's a match, it has found the source map file to use when mapping app.ts to app.js. If there is no match, then it can't bind the breakpoint, and it will turn gray.

Here are some things to try when your breakpoints turn gray:

Remote debugging

The Node.js debugger supports remote debugging for versions of Node.js >= 4.x. Specify a remote host via the address attribute. Here is an example:

{
    "type": "node",
    "request": "attach",
    "name": "Attach to remote",
    "address": "TCP/IP address of process to be debugged",
    "port": "9229"
}

By default, VS Code will stream the debugged source from the remote Node.js folder to the local VS Code and show it in a read-only editor. You can step through this code, but cannot modify it. If you want VS Code to open the editable source from your workspace instead, you can setup a mapping between the remote and local locations. A localRoot and a remoteRoot attribute can be used to map paths between a local VS Code project and a (remote) Node.js folder. This works even locally on the same system or across different operating systems. Whenever a code path needs to be converted from the remote Node.js folder to a local VS Code path, the remoteRoot path is stripped off the path and replaced by localRoot. For the reverse conversion, the localRoot path is replaced by the remoteRoot.

{
    "type": "node",
    "request": "attach",
    "name": "Attach to remote",
    "address": "TCP/IP address of process to be debugged",
    "port": "9229",
    "localRoot": "${workspaceFolder}",
    "remoteRoot": "C:\\Users\\username\\project\\server"
}

Two frequently used applications of remote debugging are:

If you are running Node.js inside a Docker container, you can use the approach from above to debug Node.js inside the Docker container and map back the remote source to files in your workspace. We have created a "recipe" on GitHub that walks you through on how to set this up Node.js in Docker with TypeScript.

If you want to run Node.js in the Windows Subsystem for Linux (WSL), you can use the approach from above as well. However, to make this even simpler, we've introduced a useWSL flag to automatically configure everything so that Node.js runs in the Linux subsystem and source is mapped to files in your workspace.

Here is the simplest debug configuration for debugging hello.js in WSL:

json { "type": "node", "request": "launch", "name": "Launch in WSL", "useWSL": true, "program": "${workspaceFolder}/hello.js" }

Please note: With the arrival of the Remote - WSL extension, VS Code got universal support for Windows Subsystem for Linux (WSL). Consequently, the useWSL debug configuration attribute has been deprecated and support for it will be dropped soon. For more details, please see our Developing in WSL documentation.

Access Loaded Scripts

If you need to set a breakpoint in a script that is not part of your workspace and therefore cannot be easily located and opened through normal VS Code file browsing, you can access the loaded scripts via the LOADED SCRIPTS view in the Run view:

Loaded Scripts Explorer

Alternatively, you can use the Debug: Open Loaded Script action (kb(extension.node-debug.pickLoadedScript)) which opens a Quick Pick, where you can filter and select the script to open.

Scripts are loaded into a read-only editor where you can set breakpoints. These breakpoints are remembered across debug sessions but you only have access to the script content while a debug session is running.

Restarting debug sessions automatically when source is edited

The restart attribute of a launch configuration controls whether the Node.js debugger automatically restarts after the debug session has ended. This feature is useful if you use nodemon to restart Node.js on file changes. Setting the launch configuration attribute restart to true makes the node debugger automatically try to reattach to Node.js after Node.js has terminated.

If you have started your program server.js via nodemon on the command line like this:

nodemon --inspect server.js

you can attach the VS Code debugger to it with the following launch configuration:

{
    "name": "Attach to node",
    "type": "node",
    "request": "attach",
    "restart": true,
    "port": 9229
}

Alternatively you can start your program server.js via nodemon directly with a launch config and attach the VS Code debugger:

{
    "name": "Launch server.js via nodemon",
    "type": "node",
    "request": "launch",
    "runtimeExecutable": "nodemon",
    "program": "${workspaceFolder}/server.js",
    "restart": true,
    "console": "integratedTerminal",
    "internalConsoleOptions": "neverOpen"
}

Tip: Pressing the Stop button stops the debug session and disconnects from Node.js, but nodemon (and Node.js) will continue to run. To stop nodemon, you will have to kill it from the command line (which is easily possible if you use the integratedTerminal as shown above).

Tip: In case of syntax errors, nodemon will not be able to start Node.js successfully until the error has been fixed. In this case, VS Code will continue trying to attach to Node.js but eventually give up (after 10 seconds). To avoid this, you can increase the timeout by adding a timeout attribute with a larger value (in milliseconds).

Automatically attach debugger to Node.js subprocesses

The Node debugger has a mechanism that tracks all subprocesses of a debuggee and tries to automatically attach to those processes that are launched in debug mode. This feature simplifies debugging of programs that fork or spawn Node.js processes like programs based on the "cluster" node module:

Auto Attach shown with Cluster Example

The feature is enabled by setting the launch config attribute autoAttachChildProcesses to true:

{
  "type": "node",
  "request": "launch",
  "name": "Cluster",
  "program": "${workspaceFolder}/cluster.js",
  "autoAttachChildProcesses": true
}

Tip: In order to be able to track the subprocesses, we need the process ID of the parent. For this we require that the main debuggee launched from the launch config is a Node.js process and we use an "evaluate" to find its process ID.

Tip: Whether a process is in debug mode is guessed by analyzing the program arguments. Currently we detect the patterns --inspect, --inspect-brk, --inspect-port, --debug, --debug-brk, --debug-port (all optionally followed by a = and a port number).

Restart frame (node)

The Node debugger supports restarting execution at a stack frame. This can be useful in situations where you have found a problem in your source code and you want to rerun a small portion of the code with modified input values. Stopping and then restarting the full debug session can be very time-consuming. The Restart Frame action allows you to reenter the current function after you have changed variables with the Set Value action:

restart frame

Note that Restart Frame won't unroll any state changes, so it may not always work as expected.

Make sure to use a Node.js version >= 5.11 since earlier versions do not work in all situations.

Breakpoints

Function breakpoints

The Node.js debugger only supports function breakpoints when the "legacy" protocol is used (that is when targeting Node.js < 8.0 versions). In addition, be aware of the following limitations when using function breakpoints:

function breakpoint

Breakpoint hit counts

The 'hit count condition' controls how many times a breakpoint needs to be hit before it will 'break' execution. The hit count syntax supported by the Node.js debugger is either an integer or one of the operators <, <=, ==, >, >=, % followed by an integer.

Some examples:

Breakpoint validation

For performance reasons, Node.js parses the functions inside JavaScript files lazily on first access. As a consequence, breakpoints don't work in source code areas that haven't been seen (parsed) by Node.js.

Since this behavior is not ideal for debugging, VS Code passes the --nolazy option to Node.js automatically. This prevents the delayed parsing and ensures that breakpoints can be validated before running the code (so they no longer "jump").

Since the --nolazy option might increase the start-up time of the debug target significantly, you can easily opt out by passing a --lazy as a runtimeArgs attribute.

When doing so, you will find that some of your breakpoints don't "stick" to the line requested but instead "jump" for the next possible line in already-parsed code. To avoid confusion, VS Code always shows breakpoints at the location where Node.js thinks the breakpoint is. In the BREAKPOINTS section, these breakpoints are shown with an arrow between requested and actual line number:

Breakpoints View

This breakpoint validation occurs when a session starts and the breakpoints are registered with Node.js, or when a session is already running and a new breakpoint is set. In this case, the breakpoint may "jump" to a different location. After Node.js has parsed all the code (for example, by running through it), breakpoints can be easily reapplied to the requested locations with the Reapply button in the BREAKPOINTS section header. This should make the breakpoints "jump back" to the requested location.

Breakpoint Actions

Skipping uninteresting code (node, chrome)

VS Code Node.js debugging has a feature to avoid source code that you don't want to step through (AKA 'Just My Code'). This feature can be enabled with the skipFiles attribute in your launch configuration. skipFiles is an array of glob patterns for script paths to skip.

For example, using:

  "skipFiles": [
    "${workspaceFolder}/node_modules/**/*.js",
    "${workspaceFolder}/lib/**/*.js"
  ]

all code in the node_modules and lib folders in your project will be skipped.

Built-in core modules of Node.js can be referred to by the 'magic name' <node_internals> in a glob pattern. The following example skips all internal modules:

  "skipFiles": [
     "<node_internals>/**/*.js"
   ]

The exact 'skipping' rules are as follows:

Skipped source is shown in a 'dimmed' style in the CALL STACK view:

Skipped source is dimmed in call stack view

Hovering over the dimmed entries explains why the stack frame is dimmed.

A context menu item on the call stack, Toggle skipping this file enables you to easily skip a file at runtime without adding it to your launch config. This option only persists for the current debugging session. You can also use it to stop skipping a file that is skipped by the skipFiles option in your launch config.

Note: The legacy protocol debugger supports negative glob patterns, but they must follow a positive pattern: positive patterns add to the set of skipped files, while negative patterns subtract from that set.

In the following (legacy protocol-only) example all but a 'math' module is skipped:

"skipFiles": [
    "${workspaceFolder}/node_modules/**/*.js",
    "!${workspaceFolder}/node_modules/math/**/*.js"
]

Note: The legacy protocol debugger has to emulate the skipFiles feature because the V8 Debugger Protocol does not support it natively. This might result in slow stepping performance.

Supported Node-like runtimes

Since the VS Code Node.js debugger communicates to the Node.js runtimes through wire protocols, the set of supported runtimes is determined by all runtimes supporting the wire protocols.

Today two wire protocols exist:

Currently these protocols are supported by specific version ranges of the following runtimes:

Runtime 'Legacy' Protocol 'Inspector' Protocol
io.js all no
Node.js < 8.x >= 6.3 (Windows: >= 6.9)
Electron < 1.7.4 >= 1.7.4
Chakra all not yet

Although it appears to be possible that the VS Code Node.js debugger picks the best protocol always automatically, we've decided for a 'pessimistic approach' with an explicit launch configuration attribute protocol and the following values:

Starting with VS Code 1.11, the default value for the protocol attribute is auto.

If your runtime supports both protocols, here are a few additional reasons for using the inspector protocol over legacy:

We try to keep feature parity between both protocol implementations but this becomes more and more difficult because the technology underlying legacy is deprecated whereas the new inspector evolves quickly. For this reason, we specify the supported protocols if a feature is not supported by both legacy and inspector.

Next steps

In case you didn't already read the Node.js section, take a look at:

To see tutorials on the basics of Node.js debugging, check out these videos:

To learn about VS Code's task running support, go to:

To write your own debugger extension, visit:

Common questions

Yes, if you've created symlinks for folders inside your project, such as with npm link, you can debug the symlinked sources by telling the Node.js runtime to preserve symlinked paths. Use the node.exe --preserve-symlinks switch in your launch configuration runtimeArgs attribute. runtimeArgs, an array of strings, are passed to the debugging session runtime executable, which defaults to node.exe.

{
    "runtimeArgs": [
        "--preserve-symlinks"
    ]
}

If your main script is inside a symlinked path, then you will also need to add the "--preserve-symlinks-main" option. This option is only available in Node 10+.

How do I debug ECMAScript modules?

If you use esm or pass --experimental-modules to Node.js in order to use ECMAScript modules, you can pass these options through the runtimeArgs attribute of launch.json: