Awesome Open Source
Awesome Open Source

OpenAPI CLI Generator

GoDoc Build Status Go Report Card Platforms


Note: this project has been superceded by Restish, an advanced auto-configured OpenAPI CLI that just works:

This project can be used to generate CLIs from OpenAPI 3 specs. The generated CLIs have the following features:

  • Authentication support for API keys and Auth0.
  • Commands, subcommands, & flag parsing through Cobra
  • Configuration through Viper
    • JSON, YAML, or TOML config files in /etc/ and $HOME, e.g. {"verbose": true} in ~/.my-app/config.json
    • From environment: APP_NAME_VERBOSE=1
    • From flags: --verbose
  • HTTP middleware through Gentleman
  • Command middleware with custom parameters (see customization below)
  • Input through stdin or CLI shorthand
  • Built-in cache to save data between runs
  • Fast structured logging via zerolog
  • Pretty output colored by Chroma
  • Response filtering & projection by JMESPath plus enhancements

Getting Started

First, make sure you have Go installed. Then, you can grab this project:

$ go get -u

Next, make your project directory and generate the commands file.

# Set up your new project
$ mkdir my-cli && cd my-cli

# Create the default main file. The app name is used for config and env settings.
$ openapi-cli-generator init <app-name>

# Generate the commands
$ openapi-cli-generator generate openapi.yaml

Last, add a line like the following to your main.go file:


If you would like to generate a client for many APIs and have each available under their own namespace, pass true instead. Next, build your client:

# Build & install the generated client.
$ go install

# Test it out!
$ my-cli --help

OpenAPI Extensions

Several extensions properties may be used to change the behavior of the CLI.

Name Description
x-cli-aliases Sets up command aliases for operations.
x-cli-description Provide an alternate description for the CLI.
x-cli-ignore Ignore this path, operation, or parameter.
x-cli-hidden Hide this path, or operation.
x-cli-name Provide an alternate name for the CLI.
x-cli-waiters Generate commands/params to wait until a certain state is reached.


The following example shows how you would set up a command that can be invoked by either list-items or simply ls:

      operationId: ListItems
        - ls


You can override the default description easily:

    description: Some info talking about HTTP headers.
    x-cli-description: Some info talking about command line arguments.


It is possible to exclude paths, operations, and/or parameters from the generated CLI. No code will be generated as they will be completely skipped.

    description: I will get included in the CLI.
    x-cli-ignore: true
    description: I will not be in the CLI :-(

Alternatively, you can have the path or operation exist in the UI but be hidden from the standard help list. Specific help is still available via my-cli my-hidden-operation --help:

    x-cli-hidden: true


You can override the default name for the API, operations, and params:

  x-cli-name: foo
    operationId: myOperation
    x-cli-name: my-op
      - name: id
        x-cli-name: item-id
        in: query

With the above, you would be able to call my-cli my-op --item-id=12.


Waiters allow you to declaratively define special commands and parameters that will cause a command to block and wait until a particular condition has been met. This is particularly useful for asyncronous operations. For example, you might submit an order and then wait for that order to have been charged successfully before continuing on.

At a high level, waiters consist of an operation and a set of matchers that select a value and compare it to an expectation. For the example above, you might call the GetOrder operation every 30 seconds until the response's JSON status field is equal to charged. Here is what that would look like in your OpenAPI YAML file:

  title: Orders API
      operationId: GetOrder
      description: Get an order's details.
        - name: id
          in: path
                type: object
                    type: string
                    enum: ['placed', 'charged', 'shipped', 'returned']
    delay: 30
    attempts: 10
    operationId: GetOrder
      - select: response.body#status
        expected: charged

The generated CLI will work like this: my-cli wait order-charged $ID where $ID corresponds to the GetOrder operation's id parameter. It will try to get and match the status 10 times, with a pause of 30 seconds between tries. If it matches, it will exit with a zero status code. If it fails, it will exit with a non-zero exit code and log a message.

This is a great start, but we can make this a little bit friendlier to use. Take a look at this modified waiter configuration:

    short: Short description for CLI `--help`
    long: Long description for CLI `--help`
    delay: 30
    attempts: 10
    operationId: GetOrder
      - select: response.body#status
        expected: charged
      - select: response.status
        expected: 404
        state: failure
        id: response.body#order_id

Here we added two new features:

  1. A short-circuit to fail fast. If we type an invalid order ID then we want the command to exit immediately with a non-zero exit code.

  2. The after block allows us to add a parameter to an existing operation to invoke the waiter. This block says that after a call to CreateOrder with a --wait-order-charged param, it should call the waiter's GetOrder operation with the id param set to the result of the response.body#order_id selector.

You can now create and wait on an order via my-cli create-order <order.json --wait-order-charged.


The following matcher fields are available:

Field Description Example
select The value selection criteria. See the selector table below. response.status
test The test to perform. Defaults to equal but can be set to any and all to match list items. equal
expected The expected value charged
state The state to set. Defaults to success but can be set to failure. success

The following selectors are available:

Selector Description Argument Example
request.param Request parameter Parameter name request.param#id
request.body Request body query JMESPath query
response.status Response HTTP status code - response.status
response.header Response HTTP header Header name response.header#content-type
response.body Response body query JMESPath query response.body#orders[].status


Your main.go is the entrypoint to your generated CLI, and may be customized to add additional logic and features. For example, you might set custom headers or handle auth before a request goes out on the wire. The apikey module provides a sample implementation.

Configuration Description

TODO: Show table describing all well-known configuration keys.

Custom Global Flags

It's possible to supply custom flags and a pre-run function. For example, say your OpenAPI spec has two servers: production and testing. You could add a --test flag to select the second server.

func main() {
	// ... init code ...

	// Add a `--test` flag to enable hitting testing.
	cli.AddGlobalFlag("test", "", "Use test endpoint", false)

	cli.PreRun = func(cmd *cobra.Command, args []string) error {
		if viper.GetBool("test") {
			// Use the test server
			viper.Set("server-index", 1)

HTTP Middleware

Gentleman provides support for HTTP request and response middleware. Don't forget to call h.Next(ctx) in your handler! For example:

// Register a request middleware handler to print the path.
cli.Client.UseRequest(func(ctx *context.Context, h context.Handler) {
	fmt.Printf("Request path: %s\n", ctx.Request.URL.Path)

// Register a response middleware handler to print the status code.
cli.Client.UseResponse(func(ctx *context.Context, h context.Handler) {
	fmt.Printf("Response status: %d\n", ctx.Response.StatusCode)

Custom Command Flags & Middleware

While the above HTTP middleware is great for adding headers or logging various things, there are times when you need to modify the behavior of a generated command. You can do so by registering custom command flags and using command middleware.

Flags and middleware are applied to a command path, which is a space-separated list of commands in a hierarchy. For example, if you have a command foo which has a subcommand bar, then the command path to reference bar for flags and middleware would be foo bar.

Note that any calls to cli.AddFlag must be made before calling the generated command registration function (e.g. openapiRegister(...)) or the flags will not get created properly.

Here's an example showing how a custom flag can change the command response:

// Register a new custom flag for the `foo` command.
cli.AddFlag("foo", "custom", "", "description", "")

cli.RegisterAfter("foo", func(cmd *cobra.Command, params *viper.Viper, resp *gentleman.Response, data interface{}) interface{} {
  m := data.(map[string]interface{})
  m["custom"] = params.GetString("custom")
  return m

// Register our generated commands with the CLI after the above.

If the foo command would normally return a JSON object like {"hello": "world"} it would now return the following if called with --custom=test:

  "custom": "test",
  "hello": "world"

Authentication & Authorization

See the apikey module for a simple example of a pre-shared key.

If instead you use a third party auth system that vends tokens and want your users to be able to log in and use the API, here's an example using Auth0:

func main() {
		AppName:   "example",
		EnvPrefix: "EXAMPLE",
		Version:   "1.0.0",

  // Auth0 requires a client ID, issuer base URL, and audience fields when
  // requesting a token. We set these up here and use the Authorization Code
  // with PKCE flow to log in, which opens a browser for the user to log into.
  clientID := "abc123"
  issuer := ""

  cli.UseAuth("user", &oauth.AuthCodeHandler{
    ClientID: "clientID",
    AuthorizeURL: issuer+"authorize",
    TokenURL: issuer+"oauth/token",
    Keys: []string{"audience"},
    Params: []string{"audience"},
    Scopes: []string{"offline_access"},

  // TODO: Register API commands here
  // ...


Note that there is a convenience module when using Auth0 specifically, allowing you to do this:

auth0.InitAuthCode(clientID, issuer,

The expanded example above is more useful when integrating with other services since it uses basic OAuth 2 primitives.


Working with Templates

The code generator is configured to bundle all necessary assets into the final executable by default. If you wish to modify the templates, you can use the go-bindata tool to help:

# One-time setup of the go-bindata tool:
$ go get -u

# Set up development mode (load data from actual files in ./templates/)
$ go-bindata -debug ./templates/...

# Now, do all your edits to the templates. You can test with:
$ go run *.go generate my-api.yaml

# Build the final static embedded files and code generator executable.
$ go generate
$ go install


Get A Weekly Email With Trending Projects For These Topics
No Spam. Unsubscribe easily at any time.
go (14,898
cli (1,763
rest-api (546
openapi (211
code-generator (108
rest-client (54