search docs

Run an actor or task and retrieve data via API

Learn how to perform the most common integration workflow - run the job => wait => collect data. Integrate Apify actors with your applications.

The most common integration of Apify with your system is usually very simple. You need to run an actor or task, wait for it to finish, then collect the data. With all the features Apify provides, new users may not be sure of the standard/easiest way to implement this. So, let's dive in and see how you can do it.

Remember to check out our API documentation with examples in different languages and live API console. We also recommend testing the API with a nice desktop client like Postman.

There are 2 ways to use the API:

Run an actor or task

API endpoints for actors and tasks and their usage (for both sync and async) are essentially the same. If you are unsure of the difference between an actor and task, read about it in the tasks documentation. In brief, tasks are just pre-configured inputs for actors.

To run (or "call" in API language) an actor/task, you will need a few things:

  • Name or ID of the actor/task. The name is in the format username~actorName or username~taskName.

  • Your API token. You can find it on the Integrations page in the Apify app (make sure it does not leak anywhere!).

  • Possibly an input or other settings if you want to change the default values (e.g. memory or build).

The template URL for a POST request to run the actor looks like this:

For tasks, we just switch the path from acts to actor-tasks:

If we send a correct POST request to this endpoint, the actor/task will start just as if we had pressed the Run button in the Apify app.

Additional settings

We can also add any settings (these will override the default settings) as additional query parameters. So if you want to change how much memory you want to allocate and which build you want to run, simply add these as parameters separated with &.

This works almost identically for actors and tasks. However, for tasks there is no reason to provide a build since a task already has only one specific actor build.

Input JSON

Most actors would not be much use if you could not pass any input to change their behavior. And even though each task already has an input, it is handy to be able to always overwrite with the API call.

An actor or task's can be any JSON object, so its structure really depends only on the specific actor. This input JSON should be sent in the POST request's body.

If you want to run one of the major actors from Apify Store, you usually do not need to provide all possible input fields. Good actors have reasonable defaults for most fields.

Let's try to run the most popular actor – the generic Web Scraper.

The full input with all possible fields is pretty long and ugly, so we will not show it here. As it has default values for most of its fields, we can provide just a simple JSON input.

We will send a POST request to the endpoint below and add the JSON as a body.

This is how it can look in Postman.

Run an actor via API in Postman

If we press Send, it will immediately return some info about the run. The status will be either READY (which means that it is waiting to be allocated on a server) or RUNNING (99% of cases).

Actor run info in Postman

We will later use this run info JSON to retrieve the data. You can also get this info about the run with another call to the Get run endpoint.

Synchronous flow

If each of your runs is shorter than 5 minutes, you can use a single synchronous endpoint. The connection is held for up to 5 minutes.

If your run exceeds this time limit, the response will be a run object containing information about the run and the status RUNNING. If that happens, you need to restart the run asynchronously and wait for the run to finish.

Synchronous runs with dataset output

Most actor runs will store their data in the default dataset. The Apify API provides run-sync-get-dataset-items endpoints for actors and tasks. These endpoints allow you to run an actor and receive the items from the default dataset.

A simple example of calling a task and logging the dataset items in Node.js.

// Use your favourite HTTP client
const got = require('got');

// Specify your API token
// (find it at
const myToken = 'rWLaYmvZeK55uatRrZib4xbZs';

// Start apify/google-search-scraper actor
// and pass some queries into the JSON body
const response = await got({
    url: `${myToken}`,
    method: 'POST',
    json: {
        queries: 'web scraping\nweb crawling'
    responseType: 'json',

const items = response.body;

// Log each non-promoted search result for both queries
items.forEach((item) => {
    const { nonPromotedSearchResults } = item;
    nonPromotedSearchResults.forEach((result) => {
        const { title, url, description } = result;
        console.log(`${title}: ${url} --- ${description}`);

Synchronous runs with key-value store output

Key-value stores are useful for storing files like images, HTML snapshots or JSON data. The Apify API provides run-sync endpoints for actors and tasks. These allow you to run a specific task and receive the output. By default, they return the OUTPUT record from the default key-value store.

For more detailed information, check the API reference.

Asynchronous flow

For runs longer than 5 minutes the process consists of three steps.

Wait for the run to finish

There may be cases where we need to simply run the actor and go away. But in any kind of integration, we are usually interested in its output. We have three basic options for how to wait for the actor/task to finish.

waitForFinish parameter

This solution is quite similar to the synchronous flow. To make the POST request wait, add the waitForFinish parameter. It can have a value from 0 to 300, which is time in seconds (maximum wait time is 5 minutes). You can extend the example URL like this:

Again, the final response will be the run info object, however now its status should be SUCCEEDED or FAILED. If the run exceeds the waitForFinish duration, the status will still be RUNNING.

You can also use waitForFinish parameter with the GET Run endpoint to implement a smarter polling system.


If you have a server, webhooks are the most elegant and flexible solution. You can simply set up a webhook for any actor or task, and that webhook sends a POST request to your server after an event occurs.

Usually, this event is a successfully finished run, but you can also set a different webhook for failed runs, etc.

Webhook example

The webhook will send you a pretty complicated JSON, but usually you are only interested in the resource object. It is essentially the run info JSON from the previous sections. You can leave the payload template as is as for our use case, since it is what we need.

Once you receive this request from the webhook, you know the event happened and you can ask for the complete data. Do not forget to respond to the webhook with a 200 status. Otherwise, it will ping you again.


There are cases where you do not have a server and the run is too long to use a synchronous call. In these cases, periodic polling of the run status is the solution.

You run the actor with the usual API call shown above. This will run the actor and give you back the run info JSON. From this JSON, extract the id field. It is the ID of the actor run that you just started.

Then, you can set an interval that will poll the Apify API (let's say every 5 seconds). In every interval, you will call the Get run endpoint to retrieve the run's status. Simply replace the RUN_ID with the ID you extracted earlier in the following URL.

Once you receive a status of SUCCEEDED or FAILED, you know the run has finished and you can cancel the interval and collect the data.

Collect the data

Unless you have used the synchronous call mentioned above, you will have to make one additional request to the API to retrieve the data.

The run info JSON also contains IDs of the default dataset and key-value store that are allocated separately for each run. This is usually everything you need. The fields are called defaultDatasetId and defaultKeyValueStoreId.

Retrieve a dataset

If you are scraping products or any list of items with similar fields, the dataset is your storage of choice. Do not forget, though, that dataset items are immutable: you can only add to the dataset, not change its content.

Retrieving the data is simple: Send a GET request to the Get items endpoint and pass the defaultDatasetId to the URL. For a GET request to the default dataset, no token is needed.

By default, it will return the data in JSON format with some metadata. The actual data are in the items array.

There are plenty of additional parameters that you can use. You can learn about them in the documentation. We will only mention that you can pass a format parameter that transforms the response into popular formats like CSV, XML, Excel, RSS, etc.

The items are paginated, which means you can ask only for a subset of the data. Specify this using the limit and offset parameters. There is actually an overall limit of 250,000 items that the endpoint can return per request. To retrieve more, you will need to send more requests incrementing the offset parameter.

Retrieve a key-value store

Key-value stores are mainly useful if you have a single output or any kind of files that cannot be stringified such as images or PDFs.

When you want to retrieve anything from a key-value store, the defaultKeyValueStoreId is not enough. You also need to know the name of the record you want to retrieve.

If you have a single output JSON, the convention is to return this as a record named OUTPUT to the default key-value store. To retrieve the record's content, call Get record endpoint. Again, no need for a token for simple GET requests.

If you do not know the keys (names) of the records in advance, you can retrieve just the keys with List keys endpoint.

Keep in mind that you can get a maximum of 1000 keys per request, so you will need to paginate over the keys using the exclusiveStartKey parameter if you have more than 1000 keys. To do this, after each call, take the last record key and provide it as the exclusiveStartKey parameter. You can do this until you get 0 keys back.