Hosted Service

Launch Options

Browserless let's you control how Chrome is started via query-string parameters in your puppeteer.connect's browserWSEndpoint URL, or in your REST API. This is useful in cases where you might need to start Chrome at a certain width and height, or need to use a proxy for visiting pages behind a firewall. Browserless (both hosted and self-hosted) lets you send launch arguments by specifying them as query-string parameters.

If you've launched workers in a different region, be sure to use the appropriate load-balancer on your account page for a faster connection.

Launching with a proxy

Please note that many proxies only support Chromium revisions in less than puppeteer 1.18.0. This is due to Chromium dropping support for Proxy-Authorization headers. Check your proxies documentation for how to properly authenticate, and ensure that you don't need to set a Proxy-Authorization header.


You can start Chrome and have it issues requests via a proxy by using the query-string parameter of --proxy-server=${server}. This will have Chrome issues requests through the proxy server as opposed to issuing plain network requests.

NOTE: If using an authenticated proxy, where you must provide a username and password, refer to your libraries documentation on how to handle this.

Blocking ads

Available in all accounts in browserless version 1.7.0 and beyond

You can use our built-in ad-blocker for both puppeteer.connect() and REST API calls. You'll simply need to add a query-string parameter of ?blockAds to your URL when connecting:


Or for REST API calls:

When new pages are created, browserless will inject a request-interceptor. If a domain matches a known site, we'll abort the call. This greatly reduces the number of inbound networks calls, and can significantly improve performance.

Running in Stealth

Newer versions of our docker image, and in our cloud environment, now include a stealth argument that utilizes puppeteer-extra-stealth to help ensure your requests run stealthfully. As with all of our launch options, simply connect or query our APIs with a stealth argument:


Or for REST API calls:

Caching with user-data-dir

Setting a user-data-dir is only available for Dedicated and self-hosted accounts (not on usage-based or cloud unit-based).

You can launch Chrome and specify a "user-data-dir" on each request to cache and make future sessions faster:


This can be used in conjuction with all our API's and integrations.

Keeping the browser alive

This feature is only supported in our puppeteer integration, and only on Dedicated and self-hosted deployments.

By default, browserless automatically closes chrome when the session disconnects. This generally meets most of the use-cases we've seen out there. However, there may be times when you want to keep the browser open even after your finished. You can do this with a keepalive parameter.

keepalive specifies how long browserless should keep the browser process open for possible reconnection later.

Keep chrome open for 5 minutes after disconnecting

  browserWSEndpoint: 'wss://',

Should you use keepalive in your connect call, you'll also want to ensure that browser.disconnect is called and not browser.close, as the latter will kill the Chrome process.

Once your app disconnects, either by design or in case of an interuppted session, we will keep the browser open for the keepalive specified (in the above example this would be 5 minutes). After that, regardless of any reconnected sessions, browserless will kill the chromium process.

In order to reconnect to an existing session, simply call the /session endpoint, or for Dedicated accounts use the GraphQL sessions query. This call will return all active sessions back, including a browserWSEndpoint payload you can use to re-connect puppeteer back to the browser. When combined with a trackingId this can make for a powerful integration and a lot of flexibility.

Starting Chrome with a width and height


You can also specify the dimensions Chrome should start with via the --window-size parameter. The value here should be a string of X and Y dimensions which translate to width and height (IE: '800,600').

Ignoring default flags

By default puppeteer specifies a certain number of launch flags in order to provide the best experience. In order to turn these off you'll need to tell browserless to ignore them. Be careful ignoring these args as it might cause chromium to become unstable or refuse to launch.


Setting headless

In many cases setting headless won't matter, as browserless will default to this to true. In some special circumastances it's important to set this to false. As an example, our screencast API sets this to false in order to appropriately load an extension for testing.


Overriding Puppeteer's launch args

Since puppeteer.connect differs greatly from puppeteer.launch, browserless a way to pass in certain parameters to "override" puppeteer's default launch arguments. These are specified through query-string parameters, and can be applied like so:

Ignore HTTPS Errors

// Via puppeteer.launch()
  ignoreHTTPSErrors: true,

// Via puppeteer.connect()
  browserWSEndpoint: 'wss://',
  ignoreHTTPSErrors: true,

Slow Mo

// Via puppeteer.launch()
  slowMo: 1000,

// Via puppeteer.connect()
  browserWSEndpoint: 'wss://',
  slowMo: 1000,

Ignore default args

// Via puppeteer.launch()
  ignoreDefaultArgs: true,

// Via puppeteer.connect()
  browserWSEndpoint: 'wss://',

Setting a tracking ID

Tracking ID's, and the workspace queries, are only available for Dedicated and self-hosted accounts (not usage-based or cloud unit-based).

Sometimes it's useful to associate particular sessions of browserless with some other part of your application. For those instances we have mechanism called trackingId.

When this query-string parameter is present in a REST API call, or puppeteer.connect, browserless will scope both file downloads and GET /sessions calls to reveal the trackingId of the original call. For instance:

// Connect with a "trackingId" in the browserWSEndpoint:

// Do some work, download files

Will save downloads to a sub-directory of AwENCwMHDQoHBgwKBAUODQ in the workspace. Thus, a GET /workspace call results in the following:

    created: '2019-07-15T22:11:40.157Z',
    isDirectory: false,
    name: 'data.csv',
    path: '/workspace/AwENCwMHDQoHBgwKBAUODQ/data.csv',
    size: 52,
    workspaceId: 'AwENCwMHDQoHBgwKBAUODQ',

Similarly, a GET /session call will reveal a property to identify that particular session:

    "description": "",
    "devtoolsFrontendUrl": "/devtools/inspector.html?ws=",
    "id": "6875C09959939E88D9BD30EC1FE1AAF0",
    "title": "Example Domain",
    "type": "page",
    "url": "",
    "webSocketDebuggerUrl": "ws://",
    "port": "62932",
    "trackingId": "AwENCwMHDQoHBgwKBAUODQ"

We'll be adding more behavioral and logging changes to surface this tracking-ID as it makes sense, so be sure to check back often to see how this can help you in your application development.

Overriding the session timer

By default, your sessions (REST API, puppeteer.connect and webdriver) are governed by a timeout. This is set via your account page for the hosted service, or by specifying CONNECTION_TIMEOUT in the docker run command. You can, on a per-job basis, override this.

For puppeteer.connect and REST API's, simply specify a ?timeout query-string parameter, with the value being the time in milliseconds for the session to execute:

# REST API (10-second limit):

# puppeteer.connect URL (10-second limit):

For webdriver, you must add an additional capability:

const webdriver = require('selenium-webdriver');

const chromeCapabilities =;
// Set the session to a 10-second limit
chromeCapabilities.set('browserless.timeout', 10000);

For other runtimes and selenium libraries be sure to consult your libraries documentation, or contact us

Other parameters

By default browserless looks for query-string parameters beginning with -- and passes them to Chrome upon boot. This means that most flags are supported even if they aren't documented. You can also see how we parse parameters by reading through our code here.

At this time, only certain flags and query-parameters are allowed for usage-based accounts and cloud unit-based, and are listed below:

  • --proxy-server,
  • --window-size,
  • --disable-web-security,
  • --enable-features,
  • --disable-web-security,
  • --disable-setuid-sandbox,
  • --lang,
  • --font-render-hinting,
  • --force-color-profile,
  • stealth,
  • ignoreDefaultArgs,
  • headless,
  • token,
  • blockAds,
  • ignoreHTTPSErrors,
  • slowMo,

Ready to try benefits of Browserless?