This doc describes the latest CLI version (15.19.0), as of this writing. If you’re using an older version of the CLI, you may want to check out these historical releases:

Getting Started

If you’re new to Zapier Platform CLI, we strongly recommend you to walk through the Tutorial for a more thorough introduction. If you haven’t used Zapier before, we would highly recommend that you learn the basics in our Zapier Getting Started Guide.

Requirements

All Zapier CLI integrations are run using Node.js v18.

You can develop using any version of Node you’d like, but your eventual code must be compatible with v18. If you’re using features not yet available in v18, you can transpile your code to a compatible format with Babel (or similar).

To ensure stability for our users, we strongly encourage you run tests on v18 sometime before your code reaches users. This can be done multiple ways.

Firstly, by using a CI tool (like Travis CI or Circle CI, which are free for open source projects). We provide a sample .travis.yml file in our template integrations to get you started.

Alternatively, you can change your local node version with tools such as nvm. Then you can either swap to that version with nvm use v18, or do nvm exec v18 zapier test so you can run tests without having to switch versions while developing.

Quick Setup Guide

First up is installing the CLI and setting up your auth to create a working “Zapier Example” integration. It will be private to you and visible in your live Zap Editor.

# install the CLI globally
npm install -g zapier-platform-cli

# setup auth to Zapier's platform with a deploy key
zapier login

Note: If you log into Zapier via the single sign-on (Google, Facebook, or Microsoft), you may not have a Zapier password. If that’s the case, you’ll need to generate a deploy key, go to your Zapier developer account here and create/copy a key, then run zapier login command with the —sso flag.

Your Zapier CLI should be installed and ready to go at this point. Next up, we’ll create our first integration!

# create a directory with the minimum required files
zapier init example-app

> Note: When you run `zapier init`, you'll be presented with a list of templates to start with. Pick the one that matches a feature you'll need (such as "dynamic-dropdown" for an integration with [dynamic dropdown fields](https://github.com/zapier/zapier-platform/blob/main/packages/cli/README.md#dynamic-dropdowns)), or select "minimal" for an integration with only the essentials. [View more example integrations here](https://github.com/zapier/zapier-platform/tree/main/example-apps).

# move into the new directory
cd example-app

# install all the libraries needed for your integration
npm install

Depending on the authentication method for your integration, you’ll also likely need to set your CLIENT_ID and CLIENT_SECRET as environment variables. These are the consumer key and secret in OAuth1 terminology.

# setting the environment variables on Zapier.com
$ zapier env:set 1.0.0 CLIENT_ID=1234
$ zapier env:set 1.0.0 CLIENT_SECRET=abcd

You should now have a working local integration. You can run several local commands to try it out.

# run the local tests
# the same as npm test, but adds some extra things to the environment
zapier test

Next, you’ll probably want to upload integration to Zapier itself so you can start testing live.

# push your integration to Zapier
zapier push

Go check out our full CLI reference documentation to see all the other commands!

Creating a Local Integration

Creating an Integration can be done entirely locally and they are fairly simple Node.js apps using the standard Node environment and should be completely testable. However, a local integration stays local until you zapier register.

# make your folder
mkdir zapier-example
cd zapier-example

# create the needed files from a template
zapier init . --template minimal

# install all the libraries needed for your integration
npm install

If you’d like to manage your local Integration, use these commands:

  • zapier init myapp - initialize/start a local integration project

  • zapier convert 1234 . - initialize/start from an existing integration

  • zapier scaffold resource Contact - auto-injects a new resource, trigger, etc.

  • zapier test - run the same tests as npm test

  • zapier validate - ensure your integration is valid

  • zapier describe - print some helpful information about your integration

Local Project Structure

In your integration’s folder, you should see this general recommended structure. The index.js is Zapier’s entry point to your integration. Zapier expects you to export an App definition there.

$ tree .
.
├── README.md
├── index.js
├── package.json
├── triggers
│   └── contact-by-tag.js
├── resources
│   └── Contact.js
├── test
│   ├── basic.js
│   ├── triggers.js
│   └── resources.js
├── build
│   └── build.zip
└── node_modules
    ├── ...
    └── ...

Local App Definition

The core definition of your App will look something like this, and is what your index.js should provide as the only export:

const App = {
  // both version strings are required
  version: require("./package.json").version,
  platformVersion: require("zapier-platform-core").version,

  // see "Authentication" section below
  authentication: {},

  // see "Dehydration" section below
  hydrators: {},

  // see "Making HTTP Requests" section below
  requestTemplate: {},
  beforeRequest: [],
  afterResponse: [],

  // See "Resources" section below
  resources: {},

  // See "Triggers/Searches/Creates" section below
  triggers: {},
  searches: {},
  creates: {},
};

module.exports = App;

Tip: You can use higher order functions to create any part of your App definition!

Zapier Platform Schema

The Zapier Platform Schema provides details for various schema components of a Zapier integration. Each schema outlines the required fields, data types, and configurations for features such as authentication, triggers, creates, searches, and HTTP requests. It’s essential for integration developers to refer to this schema documentation when building their Zapier integrations, as it ensures that the correct structure is followed and the necessary components are used. The detailed examples, valid configurations, and anti-examples help guide integraton developers in creating reliable and functional integrations, avoid common errors, and ensure compatibility with the Zapier platform.

Registering an Integration

Registering your Integration with Zapier is a necessary first step which only enables basic administrative functions. It should happen before zapier push which is to used to actually expose an Integration Version in the Zapier interface and editor.

# register your integration
zapier register "Zapier Example"

# list your integrations
zapier integrations

Note: This doesn’t put your integration in the editor - see the docs on pushing an Integration Version to do that!

If you’d like to manage your Integration, use these commands:

  • zapier integrations - list the integrations in Zapier you can administer

  • zapier register "Integration Title" - creates a new integration in Zapier

  • zapier link - lists and links a selected integration in Zapier to your current folder

  • zapier history - print the history of your integration

  • zapier team:add user@example.com admin - add an admin to help maintain/develop your integration

  • zapier users:add user@example.com 1.0.0 - invite a user try your integration version 1.0.0

Converting an Existing Integration

If you have an existing Zapier legacy Web Builder integration, you can use it as a template to kickstart your local integration.

# Convert an existing Web Builder integration to a CLI integration in the my-app directory
# Integration ID 1234 is from URL https://zapier.com/developer/builder/app/1234/development
zapier convert 1234 my-app

Your CLI integration will be created and you can continue working on it.

Note: There is no way to convert a CLI integration to a Web Builder integration and we do not plan on implementing this.

Introduced in v8.2.0, you are able to convert new integrations built in Zapier Platform UI to CLI.

# the --version flag is what denotes this command is interacting with a Visual Builder integration
# zapier convert <INTEGRATION> --version <INTEGRATION> <PATH>
zapier convert 1234 --version 1.0.1 my-app

Authentication

Most integrations require some sort of authentication. The Zapier platform provides core behaviors for several common authentication methods that might be used with your integration, as well as the ability to customize authentication further.

When a user authenticates to your integration through Zapier, a “connection” is created representing their authentication details. Data tied to a specific authentication connection is included in the bundle object under bundle.authData.

Basic

Useful if your integration requires two pieces of information to authenticate: username and password, which only the end user can provide. By default, Zapier will do the standard Basic authentication base64 header encoding for you (via an automatically registered middleware).

To create a new integration with basic authentication, run zapier init [your integration name] --template basic-auth. You can also review an example of that code here.

If your integration uses Basic auth with an encoded API key rather than a username and password, like Authorization: Basic APIKEYHERE:x, consider the Custom authentication method instead.

const authentication = {
  type: "basic",
  // "test" could also be a function
  test: {
    url: "https://example.com/api/accounts/me.json",
  },
  connectionLabel: "{{username}}", // Can also be a function, check digest auth below for an example
  // you can provide additional fields, but we'll provide `username`/`password` automatically
};

const App = {
  // ...
  authentication,
  // ...
};

Digest

Added in v7.4.0.

The setup and user experience of Digest Auth is identical to Basic Auth. Users provide Zapier their username and password, and Zapier handles all the nonce and quality of protection details automatically.

To create a new integration with digest authentication, run zapier init [your integration name] --template digest-auth. You can also review an example of that code here.

Limitation: Currently, MD5-sess and SHA are not implemented. Only the MD5 algorithm is supported. In addition, server nonces are not reused. That means for every z.request call, Zapier will send an additional request beforehand to get the server nonce.

const getConnectionLabel = (z, bundle) => {
  // bundle.inputData will contain what the "test" URL (or function) returns
  return bundle.inputData.username;
};

const authentication = {
  type: "digest",
  // "test" could also be a function
  test: {
    url: "https://example.com/api/accounts/me.json",
  },
  connectionLabel: getConnectionLabel,

  // you can provide additional fields, but we'll provide `username`/`password` automatically
};

const App = {
  // ...
  authentication,
  // ...
};

Custom

Custom auth is most commonly used for integrations that authenticate with API keys, although it also provides flexibility for any unusual authentication setup. You’ll likely provide some custom beforeRequest middleware or a requestTemplate (see Making HTTP Requests) to pass in data returned from the authentication process, most commonly by adding/computing needed headers.

To create a new integration with custom authentication, run zapier init [your integration name] --template custom-auth. You can also review an example of that code here.

const authentication = {
  type: "custom",
  // "test" could also be a function
  test: {
    url: "https://{{bundle.authData.subdomain}}.example.com/api/accounts/me.json",
  },
  fields: [
    {
      key: "subdomain",
      type: "string",
      required: true,
      helpText: "Found in your browsers address bar after logging in.",
    },
    {
      key: "api_key",
      type: "string",
      required: true,
      helpText: "Found on your settings page.",
    },
  ],
};

const addApiKeyToHeader = (request, z, bundle) => {
  request.headers["X-Subdomain"] = bundle.authData.subdomain;
  const basicHash = Buffer.from(`${bundle.authData.api_key}:x`).toString(
    "base64"
  );
  request.headers.Authorization = `Basic ${basicHash}`;
  return request;
};

const App = {
  // ...
  authentication,
  beforeRequest: [addApiKeyToHeader],
  // ...
};

Session

Session auth gives you the ability to exchange some user-provided data for some authentication data; for example, username and password for a session key. It can be used to implement almost any authentication method that uses that pattern - for example, alternative OAuth flows.

To create a new integration with session authentication, run zapier init [your integration name] --template session-auth. You can also review an example of that code here.

const getSessionKey = async (z, bundle) => {
  const response = await z.request({
    method: "POST",
    url: "https://example.com/api/accounts/login.json",
    body: {
      username: bundle.authData.username,
      password: bundle.authData.password,
    },
  });

  // response.throwForStatus() if you're using core v9 or older

  return {
    sessionKey: response.data.sessionKey,
    // or response.json.sessionKey if you're using core v9 and older
  };
};

const authentication = {
  type: "session",
  // "test" could also be a function
  test: {
    url: "https://example.com/api/accounts/me.json",
  },
  fields: [
    {
      key: "username",
      type: "string",
      required: true,
      helpText: "Your login username.",
    },
    {
      key: "password",
      type: "string",
      required: true,
      helpText: "Your login password.",
    },
    // For Session Auth we store `sessionKey` automatically in `bundle.authData`
    // for future use. If you need to save/use something that the user shouldn't
    // need to type/choose, add a "computed" field, like:
    // {key: 'something': type: 'string', required: false, computed: true}
    // And remember to return it in sessionConfig.perform
  ],
  sessionConfig: {
    perform: getSessionKey,
  },
};

const includeSessionKeyHeader = (request, z, bundle) => {
  if (bundle.authData.sessionKey) {
    request.headers = request.headers || {};
    request.headers["X-Session-Key"] = bundle.authData.sessionKey;
  }
  return request;
};

const App = {
  // ...
  authentication,
  beforeRequest: [includeSessionKeyHeader],
  // ...
};

For Session auth, the function that fetches the additional authentication data needed to make API calls (authentication.sessionConfig.perform) has the user-provided fields in bundle.inputData. Afterwards, bundle.authData contains the data returned by that function (usually the session key or token).

OAuth1

Added in v7.5.0.

Zapier’s OAuth1 implementation matches X (formerlly called Twitter) and Trello implementations of the 3-legged OAuth flow.

To create a new integration with OAuth1, run zapier init [your integration name] --template oauth1-trello. You can also check out oauth1-trello, oauth1-tumblr, and oauth1-twitter for working example integrations with OAuth1.

The flow works like this:

  1. Zapier makes a call to your API requesting a “request token” (also known as “temporary credentials”).

  2. Zapier sends the user to the authorization URL, defined by your integration, along with the request token.

  3. Once authorized, your website sends the user to the redirect_uri Zapier provided. Use zapier describe command to find out what it is:

  4. Zapier makes a backend call to your API to exchange the request token for an “access token” (also known as “long-lived credentials”).

  5. Zapier stores the access_token and uses it to make calls on behalf of the user.

You are required to define:

  • getRequestToken: The API call to fetch the request token

  • authorizeUrl: The authorization URL

  • getAccessToken: The API call to fetch the access token

You’ll also likely need to set your CLIENT_ID and CLIENT_SECRET as environment variables. These are the consumer key and secret in OAuth1 terminology.

# setting the environment variables on Zapier.com
$ zapier env:set 1.0.0 CLIENT_ID=1234
$ zapier env:set 1.0.0 CLIENT_SECRET=abcd

# and when running tests locally, don't forget to define them in .env or in the command!
$ CLIENT_ID=1234 CLIENT_SECRET=abcd zapier test

Your auth definition would look something like this:

const _ = require("lodash");

const authentication = {
  type: "oauth1",
  oauth1Config: {
    getRequestToken: {
      url: "https://{{bundle.inputData.subdomain}}.example.com/request-token",
      method: "POST",
      auth: {
        oauth_consumer_key: "{{process.env.CLIENT_ID}}",
        oauth_consumer_secret: "{{process.env.CLIENT_SECRET}}",

        // 'HMAC-SHA1' is used by default if not specified.
        // 'HMAC-SHA256', 'RSA-SHA1', 'PLAINTEXT' are also supported.
        oauth_signature_method: "HMAC-SHA1",
        oauth_callback: "{{bundle.inputData.redirect_uri}}",
      },
    },
    authorizeUrl: {
      url: "https://{{bundle.inputData.subdomain}}.example.com/authorize",
      params: {
        oauth_token: "{{bundle.inputData.oauth_token}}",
      },
    },
    getAccessToken: {
      url: "https://{{bundle.inputData.subdomain}}.example.com/access-token",
      method: "POST",
      auth: {
        oauth_consumer_key: "{{process.env.CLIENT_ID}}",
        oauth_consumer_secret: "{{process.env.CLIENT_SECRET}}",
        oauth_token: "{{bundle.inputData.oauth_token}}",
        oauth_token_secret: "{{bundle.inputData.oauth_token_secret}}",
        oauth_verifier: "{{bundle.inputData.oauth_verifier}}",
      },
    },
  },
  test: {
    url: "https://{{bundle.authData.subdomain}}.example.com/me",
  },
  // If you need any fields upfront, put them here
  fields: [
    { key: "subdomain", type: "string", required: true, default: "app" },
    // For OAuth1 we store `oauth_token` and `oauth_token_secret` automatically
    // in `bundle.authData` for future use. If you need to save/use something
    // that the user shouldn't need to type/choose, add a "computed" field, like:
    // {key: 'user_id': type: 'string', required: false, computed: true}
    // And remember to return it in oauth1Config.getAccessToken
  ],
};

// A middleware that is run before z.request() actually makes the request. Here we're
// adding necessary OAuth1 parameters to `auth` property of the request object.
const includeAccessToken = (req, z, bundle) => {
  if (
    bundle.authData &&
    bundle.authData.oauth_token &&
    bundle.authData.oauth_token_secret
  ) {
    // Just put your OAuth1 credentials in req.auth, Zapier will sign the request for
    // you.
    req.auth = req.auth || {};
    _.defaults(req.auth, {
      oauth_consumer_key: process.env.CLIENT_ID,
      oauth_consumer_secret: process.env.CLIENT_SECRET,
      oauth_token: bundle.authData.oauth_token,
      oauth_token_secret: bundle.authData.oauth_token_secret,
    });
  }
  return req;
};

const App = {
  // ...
  authentication,
  beforeRequest: [includeAccessToken],
  // ...
};

module.exports = App;

For OAuth1, authentication.oauth1Config.getRequestToken, authentication.oauth1Config.authorizeUrl, and authentication.oauth1Config.getAccessToken have fields like redirect_uri and the temporary credentials in bundle.inputData. After getAccessToken runs, the resulting token value(s) will be stored in bundle.authData for the connection.

Also, authentication.oauth1Config.getAccessToken has access to the additional return values in rawRequest and cleanedRequest should you need to extract other values (for example, from the query string).

OAuth2

Zapier’s OAuth2 implementation is based on the authorization_code flow, similar to GitHub and Facebook.

To create a new integration with OAuth2, run zapier init [your integration name] --template oauth2. You can also check out our working example integration.

If your integration’s OAuth2 flow uses a different grant type, such as client_credentials, try using Session auth instead.

The OAuth2 flow looks like this:

  1. Zapier sends the user to the authorization URL defined by your integration.

  2. Once authorized, your website sends the user to the redirect_uri Zapier provided. Use the zapier describe command to find out what it is:

  3. Zapier makes a backend call to your API to exchange the code for an access_token.

  4. Zapier stores the access_token and uses it to make calls on behalf of the user.

  5. (Optionally) Zapier can refresh the token if it expires.

You are required to define:

  • authorizeUrl: The authorization URL

  • getAccessToken: The API call to fetch the access token

If the access token has a limited life and you want to refresh the token when it expires, you’ll also need to define the API call to perform that refresh (refreshAccessToken). You can choose to set autoRefresh: true, as in the example integration, if you want Zapier to automatically make a call to refresh the token after receiving a 401. See Stale Authentication Credentials for more details on handling auth refresh.

You’ll also likely want to set your CLIENT_ID and CLIENT_SECRET as environment variables:

# setting the environment variables on Zapier.com
$ zapier env:set 1.0.0 CLIENT_ID=1234
$ zapier env:set 1.0.0 CLIENT_SECRET=abcd

# and when running tests locally, don't forget to define them in .env or in the command!
$ CLIENT_ID=1234 CLIENT_SECRET=abcd zapier test

Your auth definition would look something like this:

const authentication = {
  type: "oauth2",
  test: {
    url: "https://{{bundle.authData.subdomain}}.example.com/api/accounts/me.json",
  },
  // you can provide additional fields for inclusion in authData
  oauth2Config: {
    // "authorizeUrl" could also be a function returning a string url
    authorizeUrl: {
      method: "GET",
      url: "https://{{bundle.inputData.subdomain}}.example.com/api/oauth2/authorize",
      params: {
        client_id: "{{process.env.CLIENT_ID}}",
        state: "{{bundle.inputData.state}}",
        redirect_uri: "{{bundle.inputData.redirect_uri}}",
        response_type: "code",
      },
    },
    // Zapier expects a response providing {access_token: 'abcd'}
    // "getAccessToken" could also be a function returning an object
    getAccessToken: {
      method: "POST",
      url: "https://{{bundle.inputData.subdomain}}.example.com/api/v2/oauth2/token",
      body: {
        code: "{{bundle.inputData.code}}",
        client_id: "{{process.env.CLIENT_ID}}",
        client_secret: "{{process.env.CLIENT_SECRET}}",
        redirect_uri: "{{bundle.inputData.redirect_uri}}",
        grant_type: "authorization_code",
      },
      headers: {
        "Content-Type": "application/x-www-form-urlencoded",
      },
    },
    scope: "read,write",
  },
  // If you need any fields upfront, put them here
  fields: [
    { key: "subdomain", type: "string", required: true, default: "app" },
    // For OAuth2 we store `access_token` and `refresh_token` automatically
    // in `bundle.authData` for future use. If you need to save/use something
    // that the user shouldn't need to type/choose, add a "computed" field, like:
    // {key: 'user_id': type: 'string', required: false, computed: true}
    // And remember to return it in oauth2Config.getAccessToken/refreshAccessToken
  ],
};

const addBearerHeader = (request, z, bundle) => {
  if (bundle.authData && bundle.authData.access_token) {
    request.headers.Authorization = `Bearer ${bundle.authData.access_token}`;
  }
  return request;
};

const App = {
  // ...
  authentication,
  beforeRequest: [addBearerHeader],
  // ...
};

module.exports = App;

For OAuth2, authentication.oauth2Config.authorizeUrl, authentication.oauth2Config.getAccessToken, and authentication.oauth2Config.refreshAccessToken have fields like redirect_uri and state in bundle.inputData. After the code is exchanged for an access token and/or refresh token, those tokens are stored in bundle.authData for the connection.

Also, authentication.oauth2Config.getAccessToken has access to the additional return values in rawRequest and cleanedRequest should you need to extract other values (for example, from the query string).

If you define fields to collect additional details from the user, please note that client_id and client_secret are reserved keys and cannot be used as keys for input form fields.

Note: The OAuth2 state param is a standard security feature that helps ensure that authorization requests are only coming from your servers. Most OAuth clients have support for this and will send back the state query param that the user brings to your app. The Zapier Platform performs this check and this required field cannot be disabled. The state parameter is automatically generated by Zapier in the background, and can be accessed at bundle.inputData.state. Since Zapier uses the state to verify that GET requests to your redirect URL truly come from your integration, it needs to be generated by Zapier so that it can be validated later (once the user confirms that they’d like to grant Zapier permission to access their account in your app).

OAuth2 with PKCE

Added in v14.0.0.

Zapier’s OAuth2 implementation also supports PKCE. This implementation is an extension of the OAuth2 authorization_code flow described above.

To use PKCE in your OAuth2 flow, you’ll need to set the following variables:

  1. enablePkce: true

  2. getAccessToken.body to include code_verifier: "{{bundle.inputData.code_verifier}}"

The OAuth2 PKCE flow uses the same flow as OAuth2 but adds a few extra parameters:

  1. Zapier computes a code_verifier and code_challenge internally and stores the code_verifier in the Zapier bundle.

  2. Zapier sends the user to the authorization URL defined by your integration. We automatically include the computed code_challenge and code_challenge_method in the authorization request.

  3. Once authorized, your website sends the user to the redirect_uri Zapier provided.

  4. Zapier makes a call to your API to exchange the code but you must include the computed code_verifier in the request for an access_token.

  5. Zapier stores the access_token and uses it to make calls on behalf of the user.

Your auth definition would look something like this:

const authentication = {
  type: "oauth2",
  test: {
    url: "https://{{bundle.authData.subdomain}}.example.com/api/accounts/me.json",
  },
  // you can provide additional fields for inclusion in authData
  oauth2Config: {
    // "authorizeUrl" could also be a function returning a string url
    authorizeUrl: {
      method: "GET",
      url: "https://{{bundle.inputData.subdomain}}.example.com/api/oauth2/authorize",
      params: {
        client_id: "{{process.env.CLIENT_ID}}",
        state: "{{bundle.inputData.state}}",
        redirect_uri: "{{bundle.inputData.redirect_uri}}",
        response_type: "code",
      },
    },
    // Zapier expects a response providing {access_token: 'abcd'}
    // "getAccessToken" could also be a function returning an object
    getAccessToken: {
      method: "POST",
      url: "https://{{bundle.inputData.subdomain}}.example.com/api/v2/oauth2/token",
      body: {
        code: "{{bundle.inputData.code}}",
        client_id: "{{process.env.CLIENT_ID}}",
        client_secret: "{{process.env.CLIENT_SECRET}}",
        redirect_uri: "{{bundle.inputData.redirect_uri}}",
        grant_type: "authorization_code",
        code_verifier: "{{bundle.inputData.code_verifier}}", // Added for PKCE
      },
      headers: {
        "Content-Type": "application/x-www-form-urlencoded",
      },
    },
    scope: "read,write",
    enablePkce: true, // Added for PKCE
  },
  // If you need any fields upfront, put them here
  fields: [
    { key: "subdomain", type: "string", required: true, default: "app" },
    // For OAuth2 we store `access_token` and `refresh_token` automatically
    // in `bundle.authData` for future use. If you need to save/use something
    // that the user shouldn't need to type/choose, add a "computed" field, like:
    // {key: 'user_id': type: 'string', required: false, computed: true}
    // And remember to return it in oauth2Config.getAccessToken/refreshAccessToken
  ],
};

const addBearerHeader = (request, z, bundle) => {
  if (bundle.authData && bundle.authData.access_token) {
    request.headers.Authorization = `Bearer ${bundle.authData.access_token}`;
  }
  return request;
};

const App = {
  // ...
  authentication,
  beforeRequest: [addBearerHeader],
  // ...
};

module.exports = App;

The computed code_verifier uses this standard: RFC 7636 Code Verifier

The computed code_challenge uses this standard: RFC 7636 Code Challenge

Connection Label

When a user connects to your app via Zapier and a connection is created to hold the related data in bundle.authData, the connection is automatically labeled with the integration name. You also have the option of setting a connection label (connectionLabel), which can be extremely helpful to identify information like which user is connected or what instance of your app they are connected to. That way, users don’t get confused if they have multiple connections to your app.

When setting a connection label, you can use either a string with variable references (as shown in Basic Auth) or a function (as shown in Digest Auth).

When using a string, you have access to the information in bundle.authData and the information returned from the test request in bundle.inputData, all at the top level. So in Basic auth, if connectionLabel is {{username}}, that refers to the username used for authentication.

When using a function, this “hoisting” of data to the top level is skipped, and you must refer to data items by their fully qualified name, as shown in the line return bundle.inputData.username; in the Digest Auth snippet. return username; would not work in this context.

NOTE: Do not use sensitive authentication data such as passwords or API keys in the connection label. It’s visible in plain text on Zapier. The purpose of the label is to identify the connection for the user, so stick with data such as username or instance identifier that is meaningful but not sensitive.

Domain and subdomain validation

When adding a subdomain input field, commonly used in OAuth implementations, additional validation is strongly recommended to prevent a potential security vulnerability. If not taken into account, an attacker could utilize a maliciously constructed subdomain field (like attacker-domain.com/) in order to redirect OAuth connection requests to that attacker-controlled domain (because attacker-domain.com/.your-domain.com resolves to the attacker’s domain instead of the expected one).

This vulnerability presents itself when:

  • The authentication method uses pre-configured tokens or secret values (for example, OAuth v2)

  • User is able to input a domain or subdomain when authenticating within Zapier

  • Integration stores sensitive authentication details (in environment variables, for example) which are used as part of the authentication process

Taking the following steps prevents the potential for an attacker to access your integration’s sensitive authentication information, such as the OAuth client ID or secret.

  1. If your integration allows for the user to provide a domain, validate the input against an allow-list of trusted domains.

  2. If your integration allows for the user to provide a subdomain, add conditional validation for the subdomain string whenever you include the value in your OAuth HTTP requests. This change will prevent potential exploitation of the subdomain vulnerability.

  • If you’re using OAuth-based authentications, update the getAccessToken and optional refreshAccessToken configuration methods. If the integration uses shorthand HTTP requests, switching to manual HTTP requests will allow you to perform this manual subdomain validation.

Example code for handling subdomain validation:

const refreshAccessToken = async (z, bundle) => {
  // --- UPDATE: add your validation for the subdomain field before using it ---
  if (!/^[a-z0-9-]+$/.test(bundle.authData.yourSubdomainField)) {
    throw new Error(
      "Subdomain can only contain letters, numbers and dashes (-)."
    );
  }

  const response = await z.request({
    url: `https://${bundle.authData.yourSubdomainField}.mydomain.com/oauth/token`,
    method: "POST",
    body: {
      client_id: process.env.CLIENT_ID,
      client_secret: process.env.CLIENT_SECRET,
      grant_type: "refresh_token",
      refresh_token: bundle.authData.refresh_token,
      redirect_uri: bundle.inputData.redirect_uri,
    },
  });

  return {
    access_token: response.data.access_token,
    refresh_token: response.data.refresh_token,
  };
};

Resources

A resource is a representation (as a JavaScript object) of one of the REST resources of your API. Say you have a /recipes endpoint for working with recipes; you can define a recipe resource in your integration that will tell Zapier how to do create, read, and search operations on that resource.

const Recipe = {
  // `key` is the unique identifier the Zapier backend references
  key: "recipe",
  // `noun` is the user-friendly name displayed in the Zapier UI
  noun: "Recipe",
  // `list` and `create` are just a couple of the methods you can define
  list: {
    // ...
  },
  create: {
    // ...
  },
};

The quickest way to create a resource is with the zapier scaffold command:

zapier scaffold resource "Recipe"

This will generate the resource file and add the necessary statements to the index.js file to import it.

Resource Definition

A resource has a few basic properties. The first is the key, which allows Zapier to identify the resource on our backend. The second is the noun, the user-friendly name of the resource that is presented to users throughout the Zapier UI.

Check out this working example integration to see resources in action.

After those, there is a set of optional properties that tell Zapier what methods can be performed on the resource. The complete list of available methods can be found in the Resource Schema Docs. For now, let’s focus on two:

  • list - Tells Zapier how to fetch a set of this resource. This becomes a Trigger in the Zapier Editor.

  • create - Tells Zapier how to create a new instance of the resource. This becomes an Action in the Zapier Editor.

Here is a complete example of what the list method might look like

const Recipe = {
  key: "recipe",
  // ...
  list: {
    display: {
      label: "New Recipe",
      description: "Triggers when a new recipe is added.",
    },
    operation: {
      perform: {
        url: "https://example.com/recipes",
      },
    },
  },
};

The method is made up of two properties, a display and an operation. The display property (schema) holds the info needed to present the method as an available Trigger in the Zapier Editor. The operation (schema) provides the implementation to make the API call.

Adding a create method looks very similar.

const Recipe = {
  key: "recipe",
  // ...
  list: {
    // ...
  },
  create: {
    display: {
      label: "Add Recipe",
      description: "Adds a new recipe to our cookbook.",
    },
    operation: {
      perform: {
        method: "POST",
        url: "https://example.com/recipes",
        body: {
          name: "Baked Falafel",
          style: "mediterranean",
        },
      },
    },
  },
};

Every method you define on a resource Zapier converts to the appropriate Trigger, Create, or Search. Our examples above would result in an integration with a New Recipe Trigger and an Add Recipe Create.

Note the keys for the Trigger, Create, Search, and Search or Create are automatically generated (in case you want to use them in a dynamic dropdown), like: {resourceName}List, {resourceName}Create, {resourceName}Search, and {resourceName}SearchOrCreate; in the examples above, {resourceName} would be recipe.

Triggers/Searches/Creates

Triggers, Searches, and Creates are the way an integration defines what it is able to do. Triggers read data into Zapier (i.e. watch for new recipes). Searches locate individual records (find recipe by title). Creates create new records in your system (add a recipe to the catalog).

The definition for each of these follows the same structure. Here is an example of a trigger:

const App = {
  // ...
  triggers: {
    new_recipe: {
      key: "new_recipe", // uniquely identifies the trigger
      noun: "Recipe", // user-friendly word that is used to refer to the resource
      // `display` controls the presentation in the Zapier Editor
      display: {
        label: "New Recipe",
        description: "Triggers when a new recipe is added.",
      },
      // `operation` implements the API call used to fetch the data
      operation: {
        perform: {
          url: "https://example.com/recipes",
        },
      },
    },
    another_trigger: {
      // Another trigger definition...
    },
  },
};

You can find more details on the definition for each by looking at the Trigger Schema, Search Schema, and Create Schema.

To create a new integration with a premade trigger, search, or create, run zapier init [your integration name] and select from the list that appears. You can also check out our working example integrations here.

To add a trigger, search, or create to an existing integration, run zapier scaffold [trigger|search|create] [noun] to create the necessary files to your project. For example, zapier scaffold trigger post will create a new trigger called “New Post”.

Return Types

Each of the 3 types of function should return a certain data type for use by the platform. There are automated checks to let you know when you’re trying to pass the wrong type back. For reference, each expects:

MethodReturn TypeNotes
TriggerArray0 or more objects; passed to the deduper if polling
SearchArray0 or more objects. Only the first object will be returned, so if len > 1, put the best match first
CreateObjectReturn values are evaluated by isPlainObject

When a trigger function returns an empty array, the Zap will not trigger. For REST Hook triggers, this can be used to filter data if the available subscription options are not specific enough for the Zap’s needs.

Returning Line Items (Array of Objects)

In some cases, you may want to include multiple items in the data you return for Searches or Creates. To do that, return the set of items as an array of objects under a descriptive key. This may be as part of another object (like items in an invoice) or as multiple top-level items.

For example, a Create Order action returning an order with multiple items might look like this:

order = {
  name: 'Zap Zaplar',
  total_cost: 25.96,
  items: [
    { name: 'Zapier T-Shirt', unit_price: 11.99, quantity: 3, line_amount: 35.97, category: 'shirts' },
    { name: 'Orange Widget', unit_price: 7.99, quantity: 10, line_amount: 79.90, category: 'widgets' },
    { name:'Stuff', unit_price: 2.99, quantity: 7, line_amount: 20.93, category: 'stuff' },
    { name: 'Allbird Shoes', unit_price: 2.99, quantity: 7, line_amount: 20.93, category: 'shoes' },
  ],
  zip: 01002
}

While a Find Users search could return multiple items under an object key within an array, like this:

result = [{
  users: [
      { name: 'Zap Zaplar', age: 12, city: 'Columbia', region: 'Missouri' },
      { name: 'Orange Crush', age: 28, city: 'West Ocean City', region: 'Maryland' },
      { name: 'Lego Brick', age: 91, city: 'Billund', region: 'Denmark' },
    ],
  }];

A standard search would return just the inner array of users, and only the first user would be provided as a final result. Returning line items instead means that the “first result” return is the object containing all the user details within it.

Using the standard approach is recommended, because not all Zapier integrations support line items directly, so users may need to take additional actions to reformat this data for use in their Zaps. More detail on that at Use line items in Zaps. However, there are use cases where returning multiple results is helpful enough to outweigh that additional effort.

Fallback Sample

In cases where Zapier needs to show an example record to the user, but we are unable to get a live example from the API, Zapier will fallback to this hard-coded sample. This should reflect the data structure of the Trigger’s perform method, and have dummy values that we can show to any user.

,sample: {
  dummydata_field1: 'This will be compared against your perform method output'
  style: 'mediterranean'
}

Input Fields

Moved to Input Fields.

Output Fields

On each trigger, search, or create in the operation directive - you can provide an array of objects as fields under the outputFields. Output Fields are what users see when they select a field provided by your trigger, search or create to map it to another.

Output Fields are optional, but can be used to:

  • Define friendly labels for the returned fields. By default, we will humanize for example my_key as My Key.

  • Make sure that custom fields that may not be found in every live sample and - since they’re custom to the connected account - cannot be defined in the static sample, can still be mapped.

  • (Added in v15.6.0) Define what field(s) can be used to uniquely identify and deduplicate items returned by a polling trigger call.

The schema for outputFields is shared with inputFields but only these properties are relevant:

  • key - includes the field when not present in the live sample. When no label property is provided, key will be humanized and displayed as the field name.

  • label - defines the field name displayed to users.

  • type - defines the type for static sample data. A validation warning will be displayed if the static sample does not match the specified type.

  • required - defines whether the field is required in static sample data. A validation warning will be displayed if the value is true and the static sample does not contain the field.

  • primary - defines whether the field is part of the primary key for polling trigger deduplication.

Custom/Dynamic Output Fields are defined in the same way as Custom/Dynamic Input Fields.

Nested & Children (Line Item) Fields

To define an Output Field for a nested field use {{parent}}__{{key}}. For children (line item) fields use {{parent}}[]{{key}}.

const recipeOutputFields = async (z, bundle) => {
  const response = await z.request("https://example.com/api/v2/fields.json");

  // response.throwForStatus() if you're using core v9 or older

  // Should return an array like [{"key":"field_1","label":"Label for Custom Field"}]
  return response.data; // or response.json if you're on core v9 or older
};

const App = {
  // ...
  triggers: {
    new_recipe: {
      // ...
      operation: {
        perform: () => {},
        sample: {
          id: 1,
          title: "Pancake",
          author: {
            id: 1,
            name: "Amy",
          },
          ingredients: [
            {
              name: "Egg",
              amount: 1,
            },
            {
              name: "Milk",
              amount: 60,
              unit: "g",
            },
            {
              name: "Flour",
              amount: 30,
              unit: "g",
            },
          ],
        },
        // an array of objects is the simplest way
        outputFields: [
          {
            key: "id",
            label: "Recipe ID",
            type: "integer",
          },
          {
            key: "title",
            label: "Recipe Title",
            type: "string",
          },
          {
            key: "author__id",
            label: "Author User ID",
            type: "integer",
          },
          {
            key: "author__name",
            label: "Author Name",
            type: "string",
          },
          {
            key: "ingredients[]name",
            label: "Ingredient Name",
            type: "string",
          },
          {
            key: "ingredients[]amount",
            label: "Ingredient Amount",
            type: "number",
          },
          {
            key: "ingredients[]unit",
            label: "Ingredient Unit",
            type: "string",
          },
          recipeOutputFields, // provide a function inline - we'll merge the results!
        ],
      },
    },
  },
};

Buffered Create Actions

Added in v15.15.0. This feature is currently internal-only.

A Buffered Create allows you to create objects in bulk with a single or fewer API request(s). This is useful when you want to reduce the number of requests made to your server. When enabled, Zapier holds the data until the buffer reaches a size limit or a certain time has passed, then sends the buffered data using the performBuffer function you define.

To implement a Buffered Create, you define a buffer configuration object and a performBuffer function in the operation object. In the buffer config object, you specify how you want to group the buffered data using the groupedBy setting and the maximum number of items to buffer using the limit setting.

The performBuffer function should replace the perform function. Note that perform cannot be defined along with performBuffer. Check out the create action operation schema for details.

Similar to the general perform function accepting two arguments, z and bundle objects, the performBuffer function accepts z and bufferedBundle objects. They share the same z object, but the bufferedBundle object is different from the bundle object. The bufferedBundle object has an idempotency ID set at bufferedBundle.buffer[].meta.id for each object in the buffer. performBuffer would have to return the idempotency IDs to tell Zapier which objects were successfully written. Find the details about the bufferedBundle object here.

Here is an example of a Buffered Create action:

const performBuffer = async (z, bufferedBundle) => {
  // Grab the line items, preserving the order
  const rows = bufferedBundle.buffer.map(({ inputData }) => {
    return { title: inputData.title, year: inputData.year };
  });

  // Make the bulk-create API request
  const response = await z.request({
    method: "POST",
    url: "https://api.example.com/add_rows",
    body: {
      spreadsheet: bufferedBundle.groupedBy.spreadsheet,
      worksheet: bufferedBundle.groupedBy.worksheet,
      rows,
    },
  });

  // Create a matching result using the idempotency ID for each buffered invocation run.
  // The returned IDs will tell Zapier backend which items were successfully written.
  const result = {};
  bufferedBundle.buffer.forEach(({ inputData, meta }, index) => {
    let error = "";
    let outputData = {};

    // assuming request order matches response and
    // response.data = {
    //   "rows": [
    //     {"id": "12910"},
    //     {"id": "92830"},
    //     {"error": "Not Created"},
    //     ...
    //   ]
    // }
    if (response.data.rows.length > index) {
      // assuming an error is returned with an "error" key in the response data
      if (response.data.rows[index].error) {
        error = response.data.rows[index].error;
      } else {
        outputData = response.data.rows[index];
      }
    }

    // the performBuffer method must return a data just like this
    // {
    //   "idempotency ID 1": {
    //     "outputData": {"id": "12910"},
    //     "error": ""
    //   },
    //   "idempotency ID 2": {
    //     "outputData": {"id": "92830"},
    //     "error": ""
    //   },
    // "idempotency ID 3": {
    //     "outputData": {},
    //     "error": "Not Created"
    //   },
    //   ...
    // }
    result[meta.id] = { outputData, error };
  });

  return result;
};

module.exports = {
  key: "add_rows",
  noun: "Rows",
  display: {
    label: "Add Rows",
    description: "Add rows to a worksheet.",
  },
  operation: {
    buffer: {
      groupedBy: ["spreadsheet", "worksheet"],
      limit: 3,
    },
    performBuffer,
    inputFields: [
      {
        key: "spreadsheet",
        type: "string",
        required: true,
      },
      {
        key: "worksheet",
        type: "string",
        required: true,
      },
      {
        key: "title",
        type: "string",
      },
      {
        key: "year",
        type: "string",
      },
    ],
    outputFields: [{ key: "id", type: "string" }],
    sample: { id: "12345" },
  },
};

Deploying an Integration Version

An Integration Version is related to a specific Integration but is an “immutable” implementation of your integration. This makes it easy to run multiple versions for multiple users concurrently. The Integration Version is pulled from the version within the package.json. To create a new Integration Version, update the version number in that file. By default, every Integration Version is private but you can zapier promote it to production for use by over 1 million Zapier users.

# push your integration version to Zapier
zapier push

# list your versions
zapier versions

If you’d like to manage your Version, use these commands:

  • zapier versions - list the versions for the current directory’s integration

  • zapier push - push the current version of current directory’s integration & version (read from package.json)

  • zapier promote 1.0.0 - mark a version as the “production” version

  • zapier migrate 1.0.0 1.0.1 [100%] - move users between versions, regardless of deployment status

  • zapier deprecate 1.0.0 2020-06-01 - mark a version as deprecated, but let users continue to use it (we’ll email them)

  • zapier env:set 1.0.0 KEY=VALUE - set an environment variable to some value

  • zapier delete:version 1.0.0 - delete a version entirely. This is mostly for clearing out old test apps you used personally. It will fail if there are any users. You probably want deprecate instead.

  • zapier pull - pull the latest version from Zapier server. This is used in the event that Zapier made an update since your last version.

Note: To see the changes that were just pushed reflected in the browser, you have to manually refresh the browser each time you push.

Private Integration Version (default)

A simple zapier push will only create the Integration Version in your editor. No one else using Zapier can see it or use it.

Sharing an Integration Version

This is how you would share your integration with friends, co-workers or clients. This is perfect for quality assurance, testing with active users or just sharing any app you like.

# sends an email this user to let them view the integration version 1.0.0 in the UI privately
zapier users:add user@example.com 1.0.0

# sends an email this user to let them admin the integration (make changes just like you)
zapier team:add user@example.com

You can also invite anyone on the internet to your integration by using the links from zapier users:links. The link should look something like https://zapier.com/platform/public-invite/1/222dcd03aed943a8676dc80e2427a40d/. You can put this in your help docs, post it to Twitter, add it to your email campaign, etc. You can choose an invite link specific to an integration version or for the entire integration (i.e. all integration versions).

Promoting an Integration Version

Promotion is how you would share your integration with every one of the 1 million+ Zapier users. If this is your first time promoting - you may have to wait for the Zapier team to review and approve your integration.

If this isn’t the first time you’ve promoted your integration - you might have users on older versions. You can zapier migrate to either move users over (which can be dangerous if you have breaking changes). Or, you can zapier deprecate to give users some time to move over themselves.

# promote your integration version to all Zapier users
zapier promote 1.0.1

# OPTIONAL - migrate your users between one integration version to another
zapier migrate 1.0.0 1.0.1

# OR - mark the old version as deprecated
zapier deprecate 1.0.0 2020-06-01

Pulling Latest Version from Zapier

Zapier may fix bugs or add new features to your integration and release a new version. If you attempt to use zapier push and we’ve released a newer version, you will be prevented from pushing until you run zapier pull to update your local files with the latest version.

Any destructive file changes will prompt you with a confirmation dialog before continuing.

Environment

Integrations can define environment variables that are available when the integration’s code executes. They work just like environment variables defined on the command line. They are useful when you have data like an OAuth client ID and secret that you don’t want to commit to source control. Environment variables can also be used as a quick way to toggle between a staging and production environment during integration development.

It is important to note that variables are defined on a per-version basis! When you push a new version, the existing variables from the previous version are copied, so you don’t have to manually add them. However, edits made to one version’s environment will not affect the other versions.

Defining Environment Variables

To define an environment variable, use the env command:

# Will set the environment variable on Zapier.com
zapier env:set 1.0.0 MY_SECRET_VALUE=1234

You will likely also want to set the value locally for testing.

export MY_SECRET_VALUE=1234

Alternatively, we provide some extra tooling to work with an .env (or .environment, see below note) that looks like this:

MY_SECRET_VALUE=1234

.env is the new recommended name for the environment file since v5.1.0. The old name .environment is deprecated but will continue to work for backward compatibility.

And then in your test/basic.js file:

const zapier = require("zapier-platform-core");

should("some tests", () => {
  zapier.tools.env.inject(); // testing only!
  console.log(process.env.MY_SECRET_VALUE);
  // should print '1234'
});

This is a popular way to provide process.env.ACCESS_TOKEN || bundle.authData.access_token for convenient testing.

NOTE Variables defined via zapier env:set will always be uppercased. For example, you would access the variable defined by zapier env:set 1.0.0 foo_bar=1234 with process.env.FOO_BAR.

Accessing Environment Variables

To view existing environment variables, use the env command.

# Will print a table listing the variables for this version
zapier env:get 1.0.0

Within your integration, you can access the environment via the standard process.env - any values set via local export or zapier env:set will be there.

For example, you can access the process.env in your perform functions and in templates:

const listExample = async (z, bundle) => {
  const httpOptions = {
    headers: {
      "my-header": process.env.MY_SECRET_VALUE,
    },
  };
  const response = await z.request(
    "https://example.com/api/v2/recipes.json",
    httpOptions
  );

  // response.throwForStatus() if you're using core v9 or older

  return response.data; // or response.json if you're using core v9 or older
};

const App = {
  // ...
  triggers: {
    example: {
      noun: "{{process.env.MY_NOUN}}",
      operation: {
        // ...
        perform: listExample,
      },
    },
  },
};

Note! Be sure to lazily access your environment variables - see When to use placeholders or curlies?.

Adding Throttle Configuration

Added in v15.4.0.

When a throttle configuration is set for an action, Zapier uses it to apply throttling when the limit for the timeframe window is exceeded. It can be set at the root level and/or on an action. When set at the root level, it is the default throttle configuration used on each action of the integration. And when set in an action’s operation object, the root-level default is overwritten for that action only. Note that the throttle limit is not shared across actions unless for those with the same key, window, limit, and scope when “action” is not in the scope.

To throttle an action, you need to set a throttle object with the following variables:

  1. window [integer]: The timeframe, in seconds, within which the system tracks the number of invocations for an action. The number of invocations begins at zero at the start of each window.

  2. limit [integer]: The maximum number of invocations for an action, allowed within the timeframe window.

  3. key [string] (added in v15.6.0): The key to throttle with in combination with the scope. User data provided for the input fields can be used in the key with the use of the curly braces referencing. For example, to access the user data provided for the input field “test_field”, use {{bundle.inputData.test_field}}. Note that a required input field should be referenced to get user data always.

  4. retry [boolean] (added in v15.8.0): The effect of throttling on the tasks of the action. true means throttled tasks are automatically retried after some delay, while false means tasks are held without retry. It defaults to true. NOTE that it has no effect on polling triggers and should not be set.

  5. filter [string] (added in v15.8.0): EXPERIMENTAL: Account-based attribute to override the throttle by. You can set to one of the following: “free”, “trial”, “paid”. Therefore, the throttle scope would be automatically set to “account” and ONLY the accounts based on the specified filter will have their requests throttled based on the throttle overrides while the rest are throttled based on the original configuration.

  6. scope [array]: The granularity to throttle by. You can set the scope to one or more of the following options;

    • ‘user’ - Throttles based on user ids.

    • ‘auth’ - Throttles based on auth ids.

    • ‘account’ - Throttles based on account ids for all users under a single account.

    • ‘action’ - Throttles the action it is set on separately from other actions.

  7. overrides [array[object]] (added in v15.6.0): EXPERIMENTAL: Overrides the original throttle configuration based on a Zapier account attribute;

    • window [integer]: Same description as above.

    • limit [integer]: Same description as above.

    • filter [string]: Account-based attribute to override the throttle by. You can set to one of the following: “free”, “trial”, “paid”. Therefore, the throttle scope would be automatically set to “account” and ONLY the accounts based on the specified filter will have their requests throttled based on the throttle overrides while the rest are throttled based on the original configuration.

    • retry [boolean] (added in v15.6.1): The effect of throttling on the tasks of the action. true means throttled tasks are automatically retried after some delay, while false means tasks are held without retry. It defaults to true. NOTE that it has no effect on polling triggers and should not be set.

Both window and limit are required and others are optional. By default, throttling is scoped to the action and account.

Here is a typical usage of the throttle configuration:

const App = {
  version: require("./package.json").version,
  platformVersion: require("zapier-platform-core").version,

  // default throttle used for each action
  throttle: {
    window: 600,
    limit: 50,
    scope: ["account"],
  },

  creates: {
    upload_video: {
      noun: "Video",
      display: {
        label: "Upload Video",
        description: "Upload a video.",
      },
      operation: {
        perform: () => {},
        inputFields: [{ key: "name", required: true, type: "string" }],
        // overwrites the default, for this action
        throttle: {
          window: 600,
          limit: 5,
          key: "test-key-{{bundle.inputData.name}}",
          retry: false,
          scope: ["account"],
          overrides: [
            {
              window: 600,
              limit: 10,
              filter: "free",
              retry: false,
            },
            {
              window: 600,
              limit: 100,
              filter: "trial",
              retry: false,
            },
            {
              window: 0,
              limit: 0,
              filter: "paid",
              retry: true,
            },
          ],
        },
      },
    },
  },
};

module.exports = App;

Making HTTP Requests

There are two ways to make HTTP requests:

  1. Shorthand HTTP Requests - Easy to use, but limits what you can control. Best for simple requests.

  2. Manual HTTP Requests - Gives you full control over the request and response.

Use these helper constructs to reduce boilerplate:

  1. requestTemplate - an object literal of HTTP request options that will be merged with every request.

  2. beforeRequest - middleware that mutates every request before it is sent.

  3. afterResponse - middleware that mutates every response before it is completed.

Note: you can install any HTTP client you like - but this is greatly discouraged as you lose automatic HTTP logging and middleware.

Shorthand HTTP Requests

For simple HTTP requests that do not require special pre- or post-processing, you can specify the HTTP request options as an object literal in your app definition.

This features:

  1. Lazy {{curly}} replacement.

  2. JSON and form body de-serialization.

  3. Automatic non-2xx error raising.

const triggerShorthandRequest = {
  url: "https://{{bundle.authData.subdomain}}.example.com/v2/api/recipes.json",
  method: "GET",
  params: {
    sort_by: "id",
    sort_order: "DESC",
  },
};

const App = {
  // ...
  triggers: {
    example: {
      // ...
      operation: {
        // ...
        perform: triggerShorthandRequest,
      },
    },
  },
};

In the URL above, {{bundle.authData.subdomain}} is automatically replaced with the live value from the bundle. If the call returns a non 2xx return code, an error is automatically raised. The response body is automatically parsed as JSON or form-encoded and returned.

An error will be raised if the response cannot be parsed as JSON or form-encoded. To use shorthand requests with other response types, add middleware that sets response.data to the parsed response.

Manual HTTP Requests

Use this when you need full control over the request/response. For example:

  1. To do processing (usually involving bundle.inputData) before a request is made

  2. To do processing of an API’s response before you return data to Zapier

  3. To process an unusual response type, such as XML

To make a manual request, pass your request options to z.request() then use the resulting response object to return the data you want:

const listRecipes = async (z, bundle) => {
  // Custom processing of bundle.inputData would go here...

  const httpRequestOptions = {
    url: "https://{{bundle.authData.subdomain}}.example.com/v2/api/recipes.json",
    method: "GET",
    params: {
      cuisine: bundle.inputData.cuisine,
    },
  };
  const response = await z.request(httpRequestOptions);
  const recipes = response.data;

  // Custom processing of recipes would go here...

  return recipes;
};

const App = {
  // ...
  triggers: {
    example: {
      // ...
      operation: {
        // ...
        perform: listRecipes,
      },
    },
  },
};

Manual requests perform lazy {{curly}} replacement. In the URL above, {{bundle.authData.subdomain}} is automatically replaced with the live value from the bundle.

POST and PUT Requests

To POST or PUT data to your API you can do this:

const App = {
  // ...
  triggers: {
    example: {
      // ...
      operation: {
        // ...
        perform: async (z, bundle) => {
          const recipe = {
            name: "Baked Falafel",
            style: "mediterranean",
            directions: "Get some dough....",
          };

          const options = {
            method: "POST",
            url: "https://example.com/api/v2/recipes.json",
            body: JSON.stringify(recipe),
          };
          const response = await z.request(options);

          // Throw and try to extract message from standard error responses
          if (response.status !== 201) {
            throw new z.errors.Error(
              `Unexpected status code ${response.status}`,
              "CreateRecipeError",
              response.status
            );
          }

          return response.data; // or response.json if you're using core v9 or older
        },
      },
    },
  },
};

Note: you need to call z.JSON.stringify() before setting the body.

Using HTTP middleware

HTTP middleware is a function or piece of code that sits between a client request and the server response, allowing you to inspect, modify, or handle the request or response before they reach their destination. You use middleware to perform common tasks like adding security headers, logging requests, handling errors, or modifying data in a centralized way without repeating code. Common examples include adding a header to all outgoing responses to improve security, or catching and handling weird errors so that users receive a friendly error message instead of the system-generated message.

To process all HTTP requests in a certain way, use the beforeRequest and afterResponse middleware functions.

Middleware functions go in your app definition:

const addHeader = (request, z, bundle) => {
  request.headers["my-header"] = "from zapier";
  return request;
};

// This example only works on core v10+!
const parseXML = (response, z, bundle) => {
  // Parse content that is not JSON
  // eslint-disable-next-line no-undef
  response.data = xml.parse(response.content);
  return response;
};

// This example only works on core v10+!
const handleWeirdErrors = (response, z) => {
  // Prevent `throwForStatus` from throwing for a certain status.
  if (response.status === 456) {
    response.skipThrowForStatus = true;
  } else if (response.status === 200 && response.data.success === false) {
    throw new z.errors.Error(response.data.message, response.data.code);
  }
  return response;
};

const App = {
  // ...
  beforeRequest: [addHeader],
  afterResponse: [parseXML, handleWeirdErrors],
  // ...
};

A beforeRequest middleware function takes a request options object, and returns a (possibly mutated) request object. An afterResponse middleware function takes a response object, and returns a (possibly mutated) response object. Middleware functions are executed in the order specified in the app definition, and each subsequent middleware receives the request or response object returned by the previous middleware.

Middleware functions can be asynchronous - just return a promise from the middleware function.

The second argument for middleware is the z object, but it does not include z.request() as using that would easily create infinite loops.

Here is the full request lifecycle when you call z.request({...}):

  1. set defaults on the request object

  2. run your beforeRequest middleware functions in order

  3. add applicable auth headers (e.g. adding Basic ... for basic auth), if applicable

  4. add request.params to request.url

  5. execute the request, store the result in response

  6. try to auto-parse response body for non-raw requests, store result in response.data

  7. log the request to Zapier’s logging server

  8. if the status code is 401, you’re using a refresh-able auth (such as oauth2 or session) and autoRefresh is true in your auth configuration, throw a RefreshAuthError. The server will attempt to refresh the authentication again and retry the whole step

  9. run your afterResponse middleware functions in order

  10. call response.throwForStatus() unless response.skipThrowForStatus is true

The resulting response object is returned from z.request().

Example Integration: check out https://github.com/zapier/zapier-platform/tree/main/example-apps/middleware for a working example integration using HTTP middleware.

Error Response Handling

Since v10.0.0, z.request() calls response.throwForStatus() before it returns a response. You can disable automatic error throwing by setting skipThrowForStatus on the request object:

// Disable automatic error throwing on the request object
const perform = async (z, bundle) => {
  const response = await z.request({
    url: "...",
    skipThrowForStatus: true,
  });
  // Now you handle error response on your own.
  // The following is equivalent to response.throwForStatus(),
  // but you have to remember to do it on every request
  if (response.status >= 400) {
    throw new z.errors.ResponseError(response);
  }
};

You can also do it in afterResponse if the API uses a status code >= 400 that should not be treated as an error.

// Don't throw an error when response status is 456
const disableAutoThrowOn456 = (response, z) => {
  if (response.status === 456) {
    response.skipThrowForStatus = true;
  }
  return response;
};
const App = {
  // ...
  afterResponse: [disableAutoThrowOn456],
  // ...
};

For developers using v9.x and below, it’s your responsibility to throw an exception for an error response. That means you should call response.throwForStatus() or throw an error yourself, likely following the z.request call.

This behavior has changed periodically across major versions, which changes how/when you have to worry about handling errors. Here’s a diagram to illustrate that:

Ensure you’re handling errors correctly for your platform version. The latest released version is 15.19.0.

HTTP Request Options

Shorthand requests and manual requests support the following HTTP options:

  • url: HTTP url, you can provide it as a separate argument (z.request(url, options)) or as part of the options object (z.request({url: url, ...})).

  • method: HTTP method, default is GET.

  • headers: request headers object, format {'header-key': 'header-value'}.

  • params: URL query params object, format {'query-key': 'query-value'}.

  • body: request body, can be a string, buffer, readable stream or plain object. When it is an object/array and the Content-Type header is application/x-www-form-urlencoded the body will be transformed to query string parameters, otherwise we’ll set the header to application/json; charset=utf-8 and JSON encode the body. Default is null.

  • allowGetBody: include body in GET requests. Set to true to enable. Default is false. Set only if required by the receiving API. See section 4.3.1 in RFC 7231.

  • json: shortcut object/array/etc. you want to JSON encode into body. Default is null.

  • form: shortcut object. you want to form encode into body. Default is null.

  • raw: set this to stream the response instead of consuming it immediately. Default is false.

  • redirect: set to manual to extract redirect headers, error to reject redirect, default is follow.

  • follow: maximum redirect count, set to 0 to not follow redirects. default is 20.

  • compress: support gzip/deflate content encoding. Set to false to disable. Default is true.

  • agent: Node.js http.Agent instance, allows custom proxy, certificate etc. Default is null.

  • timeout: request / response timeout in ms. Set to 0 to disable (OS limit still applies), timeout reset on redirect. Default is 0 (disabled).

  • signal (added in v15.14.1): enables cancelling requests via a timeout set by an AbortController. More details in node-fetch docs here. Default is null.

  • size: maximum response body size in bytes. Set to 0 to disable. Default is 0 (disabled).

  • skipThrowForStatus (added in v10.0.0): don’t call response.throwForStatus() before resolving the request with response. See HTTP Response Object.

const response = await z.request({
  url: "https://example.com",
  method: "POST",
  headers: {
    "Content-Type": "application/json",
  },
  // only provide body, json or form...
  body: { hello: "world" }, // or '{"hello": "world"}' or 'hello=world'
  json: { hello: "world" },
  form: { hello: "world" },
  // access node-fetch style response.body
  raw: false,
  redirect: "follow",
  follow: 20,
  compress: true,
  agent: null,
  timeout: 0,
  size: 0,
});

HTTP Response Object

The response object returned by z.request([url], options) supports the following fields and methods:

  • status: The response status code, i.e. 200, 404, etc.

  • content: The response content as a String. For Buffer, try options.raw = true.

  • data (added in v10.0.0): The response content as an object if the content is JSON or application/x-www-form-urlencoded (undefined otherwise).

  • headers: Response headers object. The header keys are all lower case.

  • getHeader(key): Retrieve response header, case insensitive: response.getHeader('My-Header')

  • skipThrowForStatus (added in v10.0.0): don’t call throwForStatus() before resolving the request with this response.

  • throwForStatus(): Throws an error if 400 <= statusCode < 600.

  • request: The original request options object (see above).

Additionally, if request.raw is true, the raw response has the following properties:

  • json(): Get the response content as an object, if options.raw = true and content is JSON (returns a promise). undefined in non-raw requests.

  • body: A stream available only if you provide options.raw = true.

const response = await z.request({
  // options
});

// A bunch of examples for demonstration
response.status;
response.headers["Content-Type"];
response.getHeader("content-type");
response.request; // original request options
response.throwForStatus();

if (options.raw === false) {
  // (default)
  // If you're core v10+
  response.data; // same as...
  z.JSON.parse(response.content); // or...
  querystring.parse(response.content);

  // If you're core v9 or older...
  response.json; // same as
  z.JSON.parse(response.content);
} else {
  const buf = await response.buffer();
  buf.toString();

  const text = await response.text();

  const json = await response.json();

  response.body.pipe(otherStream);
}

Dehydration

Dehydration, and its counterpart Hydration, is a tool that can lazily load data that might be otherwise expensive to retrieve aggressively.

  • Dehydration - think of this as “make a pointer”, you control the creation of pointers with z.dehydrate(func, inputData, cacheExpiration) (or z.dehydrateFile(func, inputData, cacheExpiration) for files). This usually happens in a trigger step.

  • Hydration - think of this as an automatic step that “consumes a pointer” and “returns some data”, Zapier does this automatically behind the scenes. This usually happens in an action step.

This is very common when Stashing Files - but that isn’t their only use!

The method z.dehydrate(func, inputData, cacheExpiration) has two required arguments and one optional argument:

  • func - the function to call to fetch the extra data. Can be any raw function, defined in the file doing the dehydration or imported from another part of your integration. You must also register the function in the integration’s hydrators property. Note that since v10.1.0, the maximum payload size to pass to z.dehydrate / z.dehydrateFile is 6KB.

  • inputData - this is an object that contains things like a path or id - whatever you need to load data on the other side

  • cacheExpiration (optional) - this is an integer that specifies how long in seconds the response of an hydration call with a specific inputData would be cached for. If not specified, a default of 300 seconds (5 minutes) is used. After the first hydration call, subsequent ones with identical inputData within this timeframe would use the cached response of the first hydration call until the cache expires. To workaround this cache for records triggering hydration in close succession, include a unique value in the inputData, for example a timestamp in addition to the record id. Note: It was added in v15.19.0.

Why do I need to register my functions? Because of how JavaScript works with its module system, we need an explicit handle on the function that can be accessed from the App definition without trying to “automagically” (and sometimes incorrectly) infer code locations.

Here is an example that pulls in extra data for a movie:

const getMovieDetails = async (z, bundle) => {
  const url = `https://example.com/movies/${bundle.inputData.id}.json`;
  const response = await z.request(url);

  // reponse.throwForStatus() if you're using core v9 or older

  return response.data; // or response.json if you're using core v9 or older
};

const movieList = async (z, bundle) => {
  const response = await z.request("https://example.com/movies.json");

  // response.throwForStatus() if you're using core v9 or older

  return response.data.map((movie) => {
    // so maybe /movies.json is thin content but /movies/:id.json has more
    // details we want...
    movie.details = z.dehydrate(getMovieDetails, { id: movie.id }, 3600);
    return movie;
  });
};

const App = {
  version: require("./package.json").version,
  platformVersion: require("zapier-platform-core").version,

  // don't forget to register hydrators here!
  // it can be imported from any module
  hydrators: {
    getMovieDetails,
  },

  triggers: {
    new_movie: {
      noun: "Movie",
      display: {
        label: "New Movie",
        description: "Triggers when a new Movie is added.",
      },
      operation: {
        perform: movieList,
      },
    },
  },
};

module.exports = App;

And in future steps of the Zap - if Zapier encounters a pointer as returned by z.dehydrate(func, inputData, cacheExpiration) - Zapier will tie it back to your integration and pull in the data lazily.

Why can’t I just load the data immediately? Isn’t it easier? In some cases it can be - but imagine an API that returns 100 records when polling - doing 100x GET /id.json aggressive inline HTTP calls when 99% of the time Zapier doesn’t need the data yet is wasteful.

Merging Hydrated Data

As you’ve seen, the usual call to dehydrate will assign the result to an object property:

movie.details = z.dehydrate(getMovieDetails, { id: movie.id }, 3600);

In this example, all of the movie details will be located in the details property (e.g. details.releaseDate) after hydration occurs. But what if you want these results available at the top-level (e.g. releaseDate)? Zapier supports a specific keyword for this scenario:

movie.$HOIST$ = z.dehydrate(getMovieDetails, { id: movie.id }, 3600);

Using $HOIST$ as the key will signal to Zapier that the results should be merged into the object containing the $HOIST$ key. You can also use this to merge your hydrated data into a property containing “partial” data that exists before dehydration occurs:

movie.details = {
  title: movie.title,
  $HOIST$: z.dehydrate(getMovieDetails, { id: movie.id }, 3600),
};

File Dehydration

The method z.dehydrateFile(func, inputData, cacheExpiration) allows you to download a file lazily. It takes the same arguments as z.dehydrate(func, inputData, cacheExpiration) does, but is recommended when the data is a file.

An example can be found in the Stashing Files section.

What makes z.dehydrateFile different from z.dehydrate has to do with efficiency and when Zapier chooses to hydrate data. Knowing which pointers give us back files helps us delay downloading files until it’s absolutely necessary. Not only will it help avoid unnecessary file downloads, it can also prevent errors if the file has a limited availability. (Stashing files, described below, can also help with that situation.)

A good example is when users are creating Zaps in the Zap Editor. If a pointer is made by z.dehydrate, the Zap Editor will hydrate the data immediately after pulling in samples. This allows users to map fields from the hydrated data into the subsequent steps of the Zap. If, however, the pointer is made by z.dehydrateFile, the Zap Editor will wait to hydrate the file, and will display a placeholder instead. There’s nothing inside binary file data for users to map in the subsequent steps.

z.dehydrateFile was added in v7.3.0. We used to recommend using z.dehydrate for files, but we now recommend changing it to z.dehydrateFile for a better user experience.

Optional cacheExpiration argument was added in v15.19.0.

Stashing Files

It can be expensive to download and stream files or they can require complex handshakes to authorize downloads - so we provide a helpful stash routine that will take any String, Buffer or Stream and return a URL file pointer suitable for returning from triggers, searches, creates, etc.

The interface z.stashFile(bufferStringStream, [knownLength], [filename], [contentType]) takes a single required argument - the extra three arguments will be automatically populated in most cases. Here’s a full example:

const content = "Hello world!";
const url = await z.stashFile(
  content,
  content.length,
  "hello.txt",
  "text/plain"
);
z.console.log(url);
// https://zapier-dev-files.s3.amazonaws.com/cli-platform/f75e2819-05e2-41d0-b70e-9f8272f9eebf

Most likely you’d want to stream from another URL - note the usage of z.request({raw: true}):

const fileRequest = z.request({
  url: "https://example.com/file.pdf",
  raw: true,
});
const url = await z.stashFile(fileRequest); // knownLength and filename will be sniffed from the request. contentType will be binary/octet-stream
z.console.log(url);
// https://zapier-dev-files.s3.amazonaws.com/cli-platform/74bc623c-d94d-4cac-81f1-f71d7d517bc7

Note: you should only be using z.stashFile() in a hydration method or a hook trigger’s perform if you’re sending over a short-lived URL to a file. Otherwise, it can be very expensive to stash dozens of files in a polling call - for example!

See a full example with dehydration/hydration wired in correctly:

const stashPDFfunction = (z, bundle) => {
  // use standard auth to request the file
  const filePromise = z.request({
    url: bundle.inputData.downloadUrl,
    raw: true,
  });
  // and swap it for a stashed URL
  return z.stashFile(filePromise);
};

const pdfList = async (z, bundle) => {
  const response = await z.request("https://example.com/pdfs.json");

  // response.throwForStatus() if you're using core v9 or older

  // response.json.map if you're using core v9 or older
  return response.data.map((pdf) => {
    // Lazily convert a secret_download_url to a stashed url
    // zapier won't do this until we need it
    pdf.file = z.dehydrateFile(stashPDFfunction, {
      downloadUrl: pdf.secret_download_url,
    });
    delete pdf.secret_download_url;
    return pdf;
  });
};

const App = {
  version: require("./package.json").version,
  platformVersion: require("zapier-platform-core").version,

  hydrators: {
    stashPDF: stashPDFfunction,
  },

  triggers: {
    new_pdf: {
      noun: "PDF",
      display: {
        label: "New PDF",
        description: "Triggers when a new PDF is added.",
      },
      operation: {
        perform: pdfList,
      },
    },
  },
};

module.exports = App;

To create a new integration for handling files, run zapier init [your integration name] --template files. You can also check out our working example integration here.

Logging

To view the logs for your integration, use the zapier logs command.

There are three types of logs for a Zapier integration:

  • http: logged automatically by Zapier on HTTP requests

  • bundle: logged automatically on every method execution

  • console: manual logs via z.console.log() statements (see below for details)

Note that by default, this command will only fetch logs associated to your user.

For advanced logging options, including the option to fetch logs for other users or specific integration versions, look at the help for the logs command:

zapier help logs

Console Logging

To manually print a log statement in your code, use z.console.log:

z.console.log("Here are the input fields", bundle.inputData);

The z.console object has all the same methods and works just like the Node.js Console class - the only difference is we’ll log to our distributed datastore and you can view the logs via zapier logs (more below).

Viewing Console Logs

To see your z.console.log logs, do:

zapier logs --type=console

Viewing Bundle Logs

To see the bundle logs, do:

zapier logs --type=bundle

HTTP Logging

If you are using shorthand HTTP requests or the z.request() method that we provide, HTTP logging is handled automatically for you. For example:

z.request("https://57b20fb546b57d1100a3c405.mockapi.io/api/recipes").then(
  (res) => {
    // do whatever you like, this request is already getting logged! :-D
    return res;
  }
);

HTTP logging will often work with other methods of making requests as well, but if you’re using another method and having trouble seeing logs, try using z.request().

Viewing HTTP Logs

To see the HTTP logs, do:

zapier logs --type=http

To see detailed HTTP logs, including data such as headers and request and response bodies, do:

zapier logs --type=http --detailed

Error Handling

APIs are not always available. Users do not always input data correctly to formulate valid requests. Thus, it is a good idea to write integrations defensively and plan for 4xx and 5xx responses from APIs. Without proper handling, errors often have incomprehensible messages for end users, or possibly go uncaught.

Zapier provides a couple of tools to help with error handling. First is the afterResponse middleware (docs), which provides a hook for processing all responses from HTTP calls. Second is response.throwForStatus() (docs), which throws an error if the response status indicates an error (status >= 400). Since v10.0.0, we automatically call this method before returning the response, unless you set skipThrowForStatus on the request or response object. The last tool is the collection of errors in z.errors (docs), which control the behavior of Zaps when various kinds of errors occur.

General Errors

Errors due to a misconfiguration in a user’s Zap should be handled in your integration by throwing z.errors.Error with a user-friendly message and optional error and status code. Typically, this will be prettifying 4xx responses or APIs that return errors as 200s with a payload that describes the error.

Example: throw new z.errors.Error('Contact name is too long.', 'InvalidData', 400);

z.errors.Error was added in v9.3.0. If you’re on an older version of zapier-platform-core, throw a standard JavaScript Error instead, such as throw new Error('A user-friendly message').

A couple best practices to keep in mind:

  • Elaborate on terse messages. “not_authenticated” -> “Your API Key is invalid. Please reconnect your account.”

  • If the error calls out a specific field, surface that information to the user. “Provided data is invalid” -> “Contact name is invalid”

  • If the error provides details about why a field is invalid, add that in too! “Contact name is invalid” -> “Contact name is too long”

  • The second, optional argument should be a code that a computer could use to identify the type of error.

  • The last, optional argument should be the HTTP status code, if any.

The code and status can be used by us to provide relevant troubleshooting to the user when we communicate the error.

Note that if a Zap raises too many error messages it will be automatically turned off, so only use these if the scenario is truly an error that needs to be fixed.

Halting Execution

Any operation can be interrupted or “halted” (not success, not error, but stopped for some specific reason) with a HaltedError. You might find yourself using this error in cases where a required pre-condition is not met. For instance, in a create to add an email address to a list where duplicates are not allowed, you would want to throw a HaltedError if the Zap attempted to add a duplicate. This would indicate failure, but it would be treated as a soft failure.

Unlike throwing z.errors.Error, a Zap will never by turned off when this error is thrown (even if it is raised more often than not).

Example: throw new z.errors.HaltedError('Your reason.');

Stale Authentication Credentials

For integrations that require manual refresh of authorization on a regular basis, Zapier provides a mechanism to notify users of expired credentials. With the ExpiredAuthError, the current operation is interrupted and a predefined email is sent out asking the user to refresh the credentials. While the auth is disconnected, Zap runs will not be executed, to prevent more calls with expired credentials. (The runs will be Held, and the user will be able to replay them after reconnecting.)

Example: throw new z.errors.ExpiredAuthError('You must manually reconnect this auth.');

For integrations that use OAuth2 with autoRefresh: true or Session Auth, core injects a built-in afterResponse middleware that throws an error when the response status is 401. The error will signal Zapier to refresh the credentials and then retry the failed operation. You can also throw this error manually if your server doesn’t use the 401 status or you want to trigger an auth refresh even if the credentials aren’t stale.

Example: throw new z.errors.RefreshAuthError();

Handling Throttled Requests

Since v11.2.0, there are two types of errors that can cause Zapier to throttle an operation and retry at a later time. This is useful if the API you’re interfacing with reports it is receiving too many requests, often indicated by receiving a response status code of 429.

If a response receives a status code of 429 and is not caught, Zapier will re-attempt the operation after a delay. The delay can be customized by the server response containing a specific Retry-After header in your error response or with a specified time delay in seconds using a ThrottledError:

const yourAfterResponse = (resp) => {
  if (resp.status === 429) {
    throw new z.errors.ThrottledError("message here", 60); // Zapier will retry in 60 seconds
  }
  return resp;
};

Instead of a user’s Zap erroring and halting, the request will be repeatedly retried at the specified time.

For throttled requests that occur during processing of a webhook trigger’s perform, before results are produced; there is a max retry delay of 300 seconds and a default delay of 60 seconds if none is specified. For webhook processing only, if a request during the retry attempt is also throttled, it will not be re-attempted again.

Testing

There are several ways to test your Zapier integration:

  • You can use the zapier invoke command to invoke a trigger, search, create, or an auth operation locally.

  • You can write unit tests for your Zapier integration that run locally, outside of the Zapier editor.

  • You can run these tests in a CI tool like Travis.

Using zapier invoke Command

Added in v15.17.0.

The zapier invoke <ACTION_TYPE> <ACTION_KEY> CLI command emulates how the Zapier production environment would invoke your integration. Since it runs code locally, it’s a fast way to debug and test interactively without needing to deploy the code to Zapier.

Its general execution flow involves calling operation.inputFields of an action, resolving the input data to the expected types, and then calling the operation.perform method.

zapier invoke --help should be self-explanatory, but here’s a quick rundown:

# Initialize auth data in .env file
zapier invoke auth start

# Refresh auth data (for OAuth2 or Session auth)
zapier invoke auth refresh

# Test your auth data in .env
zapier invoke auth test
zapier invoke auth label

# Invoke a polling trigger
zapier invoke trigger new_recipe

# Invoke a create action
zapier invoke create add_recipe --inputData '{"name": "Pancakes"}'
zapier invoke create add_recipe --inputData @file.json

Writing Unit Tests

From v10 of zapier-platform-cli, we recommend using the Jest testing framework. After running zapier init you should find an example test to start from in the test directory.

Note: On v9, the recommendation was Mocha. You can still use Mocha if you prefer.

/* globals describe, expect, test */

const zapier = require("zapier-platform-core");

// createAppTester() makes it easier to test your integration. It takes your raw app
// definition, and returns a function that will test you integration.
const App = require("../index");
const appTester = zapier.createAppTester(App);

// Inject the vars from the .env file to process.env. Do this if you have a .env
// file.
zapier.tools.env.inject();

describe("triggers", () => {
  test("new recipe", async () => {
    const bundle = {
      inputData: {
        style: "mediterranean",
      },
    };

    const results = await appTester(
      App.triggers.recipe.operation.perform,
      bundle
    );
    expect(results.length).toBeGreaterThan(1);

    const firstRecipe = results[0];
    expect(firstRecipe.id).toBe(1);
    expect(firstRecipe.name).toBe("Baked Falafel");
  });
});

Using the z Object in Tests

Introduced in core@11.1.0, appTester can now run arbitrary functions:

/* globals describe, expect, test */

const zapier = require("zapier-platform-core");

const App = require("../index");
const appTester = zapier.createAppTester(App);

describe("triggers", () => {
  test("new recipe", async () => {
    const adHocResult = await appTester(
      // your in-line function takes the same [z, bundle] arguments as normal
      async (z, bundle) => {
        // requests are made using your integration's actual middleware
        // make sure to pass the normal `bundle` arg to `appTester` if your requests need auth
        const response = await z.request(
          "https://example.com/some/setup/method",
          {
            params: {
              numItems: bundle.inputData.someValue,
            },
          }
        );

        return {
          // you can use all the functions on the `z` object
          someHash: z.hash("md5", "mySecret"),
          data: response.data,
        };
      },
      {
        // you must provide auth data for authenticated requests
        // (just like running a normal trigger)
        authData: { token: "some-api-key" },
        // put arbitrary function params in `inputData`
        inputData: {
          someValue: 3,
        },
      }
    );

    expect(adHocResult.someHash).toEqual("a5beb6624e092adf7be31176c3079e64");
    expect(adHocResult.data).toEqual({ whatever: true });

    // ... rest of test
  });
});

Mocking Requests

It’s useful to test your code without actually hitting any external services. Nock is a Node.js utility that intercepts requests before they ever leave your computer. You can specify a response code, body, headers, and more. It works out of the box with z.request by setting up your nock before calling appTester.

/* globals describe, expect, test */

const zapier = require("zapier-platform-core");

const App = require("../index");
const appTester = zapier.createAppTester(App);

const nock = require("nock");

describe("triggers", () => {
  test("new recipe", async () => {
    const bundle = {
      inputData: {
        style: "mediterranean",
      },
    };

    // mocks the next request that matches this url and querystring
    nock("https://example.com/api")
      .get("/recipes")
      .query(bundle.inputData)
      .reply(200, [
        { name: "name 1", directions: "directions 1", id: 1 },
        { name: "name 2", directions: "directions 2", id: 2 },
      ]);

    const results = await appTester(
      App.triggers.recipe.operation.perform,
      bundle
    );

    expect(results.length).toBeGreaterThan(1);

    const firstRecipe = results[0];
    expect(firstRecipe.id).toBe(1);
    expect(firstRecipe.name).toBe("name 1");
  });
});

Here’s more info about nock and its usage in the README.

Running Unit Tests

To run all your tests do:

zapier test

You can also go direct with npm test or node_modules/.bin/jest.

Testing & Environment Variables

The best way to store sensitive values (like API keys, OAuth secrets, or passwords) is in an .env (or .environment, see below note) file (learn more). Then, you can include the following before your tests run:

const zapier = require("zapier-platform-core");
zapier.tools.env.inject(); // inject() can take a filename; defaults to ".env"

// now process.env has all the values in your .env file

.env is the new recommended name for the environment file since v5.1.0. The old name .environment is deprecated but will continue to work for backward compatibility.

Remember: NEVER add your secrets file to version control!

Additionally, you can provide them dynamically at runtime:

CLIENT_ID=1234 CLIENT_SECRET=abcd zapier test

Or, export them explicitly and place them into the environment:

export CLIENT_ID=1234
export CLIENT_SECRET=abcd
zapier test

Testing in Your CI

Whether you use Travis, Circle, Jenkins, or another service, we aim to make it painless to test in an automated environment.

Behind the scenes zapier test does a standard npm test, which could be Jest or Mocha, based on your project setup.

This makes it straightforward to integrate into your testing interface. For example, if you want to test with Travis CI, the .travis.yml would look something like this:

language: node_js
node_js:
  - "v18"
before_script: npm install -g zapier-platform-cli
script: CLIENT_ID=1234 CLIENT_SECRET=abcd zapier test

You can substitute zapier test with npm test, or a direct call to node_modules/.bin/jest. We recommend putting environment variables directly into the configuration screens Jenkins, Travis, or other services provide.

Alternatively to reading the deploy key from root (the default location), you may set the ZAPIER_DEPLOY_KEY environment variable to run privileged commands without the human input needed for zapier login. We suggest encrypting your deploy key in the manner your CI provides (such as these instructions, for Travis).

Debugging Tests

Sometimes tests aren’t enough, and you may want to step through your code and set breakpoints. The testing suite is a regular Node.js process, so debugging it doesn’t take anything special. Because we recommend jest for testing, these instructions will outline steps for debugging w/ jest, but other test runners will work similarly. You can also refer to Jest’s own docs on the subject.

To start, add the following line to the scripts section of your package.json:

"test:debug": "node --inspect-brk node_modules/.bin/jest --runInBand"

This will tell node to inspect the jest processes, which is exactly what we need.

Next, add a debugger; statement somewhere in your code, probably in a perform method:

// triggers on a new pizza with a certain tag
const perform = async (z, bundle) => {
  const response = await z.request({
    url: "https://jsonplaceholder.typicode.com/posts",
    params: {
      tag: bundle.inputData.tagName,
    },
  });
  debugger;
  // this should return an array of objects
  return response.data;
};

This creates a breakpoint while inspecting, or a starting point for our manual inspection.

Next, you’ll need an inspection client. The most available one is probably the Google Chrome browser, but there are lots of options. We’ll use Chrome for this example. In your terminal (and in your integration’s root directory), run yarn test:debug (or npm run test:debug). You should see the following:

% yarn test:debug
yarn run v1.22.10
$ node --inspect-brk node_modules/.bin/jest --runInBand
Debugger listening on ws://127.0.0.1:9229/5edaab3c-a1d3-45e4-b374-0536095c559b
For help, see: https://nodejs.org/en/docs/inspector

Now in Chrome, go to chrome://inspect. Make sure Discover Network Targets is checked and you should see a path to your jest file on your local machine:

Click inspect. A new window will open. Next, click the little blue arrow in the top right to actually run the code:

After a few seconds, you’ll see your code, the debugger statement, and info about the current environment on the right panel. You should see familiar data in the Locals section, such as the response variable, and the z object.

Debugging combined with thorough unit tests will hopefully equip you in keeping your Zapier integration in smooth working order.

Using npm Modules

Use npm modules just like you would use them in any other node app, for example:

npm install --save jwt

And then package.json will be updated, and you can use them like anything else:

const jwt = require("jwt");

During the zapier build or zapier push step - we’ll copy all your code to a temporary folder and do a fresh re-install of modules.

Note: If your package isn’t being pushed correctly (IE: you get “Error: Cannot find module ‘whatever’” in production), try adding the --disable-dependency-detection flag to zapier push.

Note 2: You can also try adding a includeInBuild array property (with paths to include, which will be evaluated to RegExp, with a case insensitive flag) to your .zapierapprc file, to make it look like:

{
  "id": 1,
  "key": "App1",
  "includeInBuild": ["test.txt", "testing.json"]
}

Warning: Do not use compiled libraries unless you run your build on the AWS AMI ami-4fffc834, or follow the Docker instructions below.

Building Native Packages with Docker

Unfortunately if you are developing on a macOS or Windows box you won’t be able to build native libraries locally. If you try and push locally build native modules, you’ll get runtime errors during usage. However, you can use Docker and Docker Compose to do this in a pinch. Make sure you have all the necessary Docker programs installed and follow along.

First, create your Dockerfile:

FROM amazonlinux:2017.03.1.20170812

RUN yum install zip findutils wget gcc44 gcc-c++ libgcc44 cmake -y

RUN wget https://nodejs.org/dist/v8.10.0/node-v8.10.0.tar.gz && \
    tar -zxvf node-v8.10.0.tar.gz && \
    cd node-v8.10.0 && \
    ./configure && \
    make && \
    make install && \
    cd .. && \
    rm -rf node-v8.10.0 node-v8.10.0.tar.gz

RUN npm i -g zapier-platform-cli

WORKDIR /app

And finally, create your docker-compose.yml file:

version: "3.4"

services:
  pusher:
    build: .
    volumes:
      - .:/app
      - node_modules:/app/node_modules:delegated
      - ~/.zapierrc:/root/.zapierrc
    command: 'bash -c "npm i && zapier push"'
    environment:
      ZAPIER_DEPLOY_KEY: ${ZAPIER_DEPLOY_KEY}

volumes:
  node_modules:

Note: Watch out for your package-lock.json file, if it exists for local install it might incorrectly pin a native version.

Now you should be able to run docker-compose run pusher and see the build and push successfully complete!

Using Transpilers

If you would like to use a transpiler like babel, you can add a script named _zapier-build to your package.json, which will be run during zapier build, zapier push, and zapier upload. See the following example:

{
  "scripts": {
    "zapier-dev": "babel src --out-dir lib --watch",
    "_zapier-build": "babel src --out-dir lib"
  }
}

Then, you can have your fancy ES7 code in src/* and a root index.js like this:

module.exports = require("./lib");

And work with commands like this:

# watch and recompile
npm run zapier-dev

# tests should work fine
zapier test

# every build ensures a fresh build
zapier push

There are a lot of details left out - check out the full example integration here.

We recommend using zapier init . to create an integration - you’ll be presented with a list of currently available example templates to start with.

Command Line Tab Completion

Introduced in v9.1.0, the zapier autocomplete command shows instructions for generating command line autocomplete.

Follow those instructions to enable completion for zapier commands and flags!

The Zapier Platform Packages

The Zapier Platform consists of 3 npm packages that are released simultaneously.

  • zapier-platform-cli is the code that powers the zapier command. You use it most commonly with the test, scaffold, and push commands. It’s installed with npm install -g zapier-platform-cli and does not correspond to a particular integration.

  • zapier-platform-core is what allows your integration to interact with Zapier. It holds the z object and integration tester code. Your integration depends on a specific version of zapier-platform-core in the package.json file. It’s installed via npm install along with the rest of your integrations’s dependencies.

  • zapier-platform-schema enforces integration structure behind the scenes. It’s a dependency of core, so it will be installed automatically.

To learn more about the structure of the code (especially if you’re interested in contributing), check out the ARCHITECTURE.md file(s).

Updating These Packages

The Zapier platform and its tools are under active development. While you don’t need to install every release, we release new versions because they are better than the last. We do our best to adhere to Semantic Versioning wherein we won’t break your code unless there’s a major release. Otherwise, we’re just fixing bugs (patch) and adding features (minor).

Broadly speaking, all releases will continue to work indefinitely. While you never have to upgrade your integration’s zapier-platform-core dependency, we recommend keeping an eye on the changelog to see what new features and bug fixes are available.

For more info about which Node versions are supported, see the faq.

The most recently released version of cli and core is 15.19.0. You can see the versions you’re working with by running zapier -v.

To update cli, run npm install -g zapier-platform-cli.

To update the version of core your integration depends on, set the zapier-platform-core dependency in your package.json to a version listed here and reinstall your dependencies (either yarn or npm install).

For maximum compatibility, keep the versions of cli and core in sync.