Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


docs - github - npm

"This CLI of mine, I'm gonna let it shine!"

This is a no-fuss CLI framework. It solves the CLI stuff, so you can focus on the other stuff.

What's it solve for you?

  • Parsing and validating command-line arguments with command-line-args.
  • Calling the correct function based on the command a user asked for.
  • Arbitrary nesting of subcommands using an Express-like middleware pattern.
  • Encapsulating all I/O streams for 100% testability -- assuming you use its logging primitives.
  • Converting the metadata you provide into nice-looking help text with command-line-usage.
  • Handling execution errors.
  • Determining an appropriate exit code for the process.
  • Resource-style commands (e.g. myapp [verb] [noun])

See the online docs for API reference information.

  1. Getting Started
    1. Terminology
  2. Handlers
    1. Handler Results
    2. Handler Errors
  3. Subcommands
    1. Subcommand Options
    2. Subcommands Sharing Data
    3. Required Subcommands
  4. Resources
  5. Help Text
  6. Version Text
  7. Error Strategies
  8. Testing
  9. FAQ
  10. Development

Getting Started

Install cli-of-mine from npm:

Important Note: cli-of-mine is tested with the latest versions of node v8, v10, and v12, on Ubuntu, Windows, and OS X. Any other Node versions (or OSes) are not officially supported.

npm i cli-of-mine

cli-of-mine follows strict semantic versioning according to semver v2.0.0.

Basic Usage

The cli-of-mine module exports a single function, exec. The function expects an ExecConfig object that describes how the CLI should work.

A "hello world" example just needs to define a name and a handler:

const { exec } = require("cli-of-mine");

  name: "my-app",

  handler(ctx) {
    ctx.console.log("Hello, world!");
}).then(result => process.exit(result.processExitCode));

For a full list of the possible options, see the ExecConfig object in the API documentation.


The following terms have a specific meaning when used in this documentation.

  • config - A configuration option for cli-of-mine. Provided by the developer, not the user.
  • argument - A single command-line argument, provided by the user.
  • command - An argument that matches the name of a command definition.
  • option - An argument that starts with a - and might have one or more values, e.g. --force or --file foo.txt.


When exec is called, it calls one or more Handlers based on the commands/subcommands that the user is invoking. The handlers implement the "business logic" or "application logic" for the commands.

For example, this handler prints "Hello, World!" using the cli-of-mine logging primitives.

const { exec } = require("cli-of-mine");

  name: "my-app",
  handler(ctx) {
    ctx.console.log("Hello, world!");

Handlers are functions that get called with two arguments: a context object (see HandlerContext) and a next function (see HandlerCallback).

Handler Results

Handlers can return a value of any type. This provides a way for handlers to communicate with the code that called exec.

Whatever they return will be returned to the next tier of middleware and eventually put into the result property of the ExecResult.

Also, if a handler returns a Promise, the next subcommand handler won't execute until the promise is resolved.

For example, this will cause the ExecResult.result property to be equal to ctx.args:

function handler(ctx, next) {
  return Promise.resolve(ctx.args);

Handler Errors

Any errors thrown within a handler will be caught and dealt with automatically.

Handlers can buy into improved error handling by throwing or rejecting with instances of AppError. It extends Error with support for error codes, and it automatically "namespaces" your codes by prefixing them with APP_, so they won't conflict with cli-of-mine codes (or Node codes).

For example:

const { AppError } = require("cli-of-mine");

throw new AppError(
  "SERVER_UNAVAILABLE", // error code
  "The downstream server is unavailable." // message

Additionally, you can set the optional processExitCode property to recommend an exit code to use, should this error become fatal.

You can construct an AppError from an existing error using fromError:

try {
  // do stuff
} catch (e) {
  throw AppError.fromError(e);

Custom Error Classes

It's recommended that you extend AppError to implement your own error class, so you can document error codes more cleanly in your app. For example:

class FooError extends AppError {
  code_prefix = "FOO_";

  constructor(code: string | null, message: string) {
    super(code, message);
    Error.captureStackTrace(this, FooError);

try {
  throw new FooError("MY_CODE", "my message");
} catch (e) {
  console.log(e.code); // logs FOO_MY_CODE


cli-of-mine is designed to support arbitrarily nested levels of subcommands using a middleware pattern inspired by frameworks like Express. This allows you to handle options at every subcommand level.

The special thing about middlewares is that your handler must choose when to relinquish control to the next handler by calling next() (which returns a Promise). For example:

function handler(ctx, next) {
  // things to do before subcommand starts

  // next() runs the subcommand's handler
  return next().then(result => {
    // things to do after subcommand is finished

Your final handler (the "controller" in Express parlance) can call next if it wants, but it doesn't have to. If it does, the next is a no-op.

Subcommand Options

Each "level" of command/subcommand specifies its own set of options. The handler is only provided the options relevant to that specific subcommand, not the ones before or after it.

For example, assume widgets is a subcommand of your application and list is a subcommand of widgets -- so users can run myapp widgets list.

If someone runs: myapp -v widgets list --filter green, then the options will be doled out like so:

  1. The root handler is given the -v argument.
  2. The widgets handler is given no arguments.
  3. The list handler is given the --filter green argument.

If the list handler actually cares about the -v argument, the root handler has to give it that information explicitly using (see below).

Subcommands Sharing Data

You cannot provide arguments to next().

If you want to share data between subcommands, you should assign it to the object, which is reserved for arbitrary handler data.

For example, the following handler will initialize a database connection for the subcommand to use, then clean it up once the subcommand is finished.

Note: The async/await syntax is used in this and remaining examples to improve readability. But good ol' Promise chains work just fine.

async function handler(ctx, next) {
  const { args, data } = ctx;

  // initialize dbConn before running subcommand
  data.dbConn = await getFooDatabaseConn(args.db);

  const result = await next();

  // Clean up dbConn after subcommand is done
  await data.dbConn.close();

  // make sure to return the subcommand's result if you care about it.
  return result;

Required Subcommands

Handlers have access to the subcommand property of HandlerContext, which indicates which subcommand (if any) the user has requested. This can be used, for instance, to throw an error if no subcommands are specified:

function handler(ctx, next) {
  if (!ctx.subcommand) {
    throw new AppError("BAD_COMMAND", "Must specify command");
  return next();


cli-of-mine includes support for declaring resources and verbs using the resources property of ExecConfig. This provides support for the myapp [verb] [noun] invocation pattern, as used by kubectl for instance.

As a trivial example, say we want to make a CLI that lets you run:

myapp add widget
myapp get widget
myapp rm widget

We can do this by defining a widget resource:

  name: "myapp",

  resources: [
      name: "widget",
      commands: [
          name: "add",
          handler: ctx => ctx.console.log("add widget")
          name: "get",
          handler: ctx => ctx.console.log("get widget")
          name: "rm",
          handler: ctx => ctx.console.log("rm widget")

Resources and subcommands can be used together. If a resource and subcommand overlaps, the subcommand is chosen and executed instead. This can be used to define "default" behavior for a given verb, for example:

  name: "myapp",

  resources: [
      name: "widget",
      commands: [
          name: "run",
          handler: ctx => ctx.console.log("Running widget")

  subcommands: [
      name: "run",
      handler: ctx => ctx.console.log("Default run behavior")

In this case, both these invocations are valid:

myapp run widget    # prints "Running widget"
myapp run           # prints "Default run behavior"

Help Text

cli-of-mine will intercept --help flags and handle them automatically. When this happens, none of your handlers will be called and the result of the execution will be undefined.

Different text is shown based on which command receives the --help flag. For example:

# displays root help for myapp
myapp --help

# displays help for "widgets" command
myapp widgets --help

# displays help for "list" subcommand
myapp widgets list --help

You can disable this functionality by setting generateHelp: false in your ExecConfig.

Version Text

cli-of-mine will intercept --version options (unless they are passed to a subcommand) and print version information based on the version property of the ExecConfig. This behavior can be disabled by setting generateVersion: false.

Error Strategies

exec provides three strategies for automated error handling, which you can pick using the errorStrategy property of the ExecConfig.

The "log" strategy is the default.


When errorStrategy: "log", the exec function will automatically catch and log any errors that occur during execution, including errors that your handlers throw. This means it never rejects.

The ExecResult will contain a processExitCode property that indicates what exit code is recommended to be used. exec will not exit the process automatically.

This is the default mode. It's useful if you want cli-of-mine to handle as much as possible, but you don't want it to exit the process for you.


when errorStrategy: "throw", errors during execution will cause the returned Promise to be rejected with an ExecutionError that you can inspect and handle manually.

This mode is useful for testing, or for cases where you want the code calling exec to be able to catch errors from within your handlers.


When errorStrategy: "exit", errors during execution will be logged to the user. If the execution would result in a nonzero exit code, the process will be automatically exited with that code.

This mode is useful if you want cli-of-mine to completely manage error handling, and you don't need to run any code after exec is finished.


One core design goal of cli-of-mine is to be testable from the user's perspective. In other words, it should be possible to easily write tests like:

"If I pass the --foo option, the output should include 'bar'."

This can be done by using the stdio-related parameters on ExecConfig, assuming your application uses the cli-of-mine logging primitives. For example:

  name: "testapp",
  options: [{ name: "foo" }],

  argv: ["--foo"],
  stdout: "capture",

  async handler(ctx, next) {
    const { foo } = ctx.args;
    ctx.console.log(foo ? "bar" : "bad request!");
}).then(result => {


Q: Is there an @types/cli-of-mine package?

No -- cli-of-mine is written in Typescript and includes the type definitions bundled in the package. No additional @types are needed.

Q: Why does my application have to use your special console?

cli-of-mine has a goal of "Capturing All I/O." But, it also has a strict "No Global Changes" rule and that applies to changing the global.console object.

The only way we can capture all IO without global changes, is to allow our applications to be configured with a custom logger. Therefore, cli-of-mine gives us the tools to do that.

If you don't care about global changes, your handler can assign the console for global use with:

global.console = ctx.console;

Q: What prompted you to create cli-of-mine?

When I was working on a Node CLI application, repost, one of my goals was to create an application that controls all its inputs and outputs, and has no global state or singletons.

As part of that, I wanted to make a CLI framework that allowed you to completely control your application's I/O, for instance to make assertions about output data. This is that framework.

Q: But what about all the existing CLI frameworks?

Surprisingly few CLI frameworks are designed for effective testing from the user's perspective. If I call the program with X arguments, do I see Y results? That can be a tough question to answer, even if you're using a CLI framework. The "special" thing about cli-of-mine is that it tries to make this kind of assertion easier using the stdout: "capture" option.

Having said that, the "established" Node CLI frameworks are still great, they just don't fit the grooves in my brain as well as this one does.

Q: When should I not use cli-of-mine?

  • You just want to parse arguments and you don't like inversion of control. In that case, try command-line-args or minimist instead.
  • You want all the batteries included. In that case, there are some larger frameworks like oclif and yargs that might be able to help you.
  • You want a well-supported library with low bus-factor. This is a personal project and I do not commit to long-term support. (This may change in the future.) For a similarly-featured, very popular library, try commander.
  • You use non-utf8 encodings. For now, cli-of-mine works best with utf8 encoded streams.


This section is for folks wanting to make changes to cli-of-mine itself.

It all starts with cloning the repository:

git clone

cd cli-of-mine

Then you can go ahead and run the tests:

# run tests once
npm run test

# automatically re-run tests when files change
npm run test:watch

cli-of-mine roughly hews to a test-driven development style. All features have at least one "acceptance test."

When new features or bugfixes are added, the changes must come with acceptance tests.

API Documentation

The library API documentation is generated by TypeDoc. The following documentation-related commands are available:

# rebuild /docs
npm run docs

# rebuild docs and run http server on localhost:8080
npm run docs:serve

# like docs:serve, but rebuilds when files change
npm run docs:watch

Copyright 2019 Luke Turner - Published under the MIT License.