Agent

This is the primary class to interact with SecretAgent. The following is a simple example:

const agent = require('secret-agent');

(async () => {
  await agent.goto('https://www.google.com');
  // other actions...
  await agent.close();
})();

An Agent instance can be thought of as a single user-browsing session. A default instance is automatically initialized and available as the default export of secret-agent. Each additional instance you create has the following attributes:

Replayable

An instance has a replayable Session that will record all commands, dom changes, interaction and page events.

Lightweight

Instances are very lightweight, sharing a pool of browsers underneath. To manage concurrent scrapes in a single script, you can create one Agent for each scrape, or manage load and concurrency with a Handler.

Single Active Tab

Agent instances can have multiple Tabs, but only a single tab can be focused at a time. Clicks and other user interaction will go to the active tab (interacting with multiple tabs at once by a single user is easily detectable).

Sandboxed

Each Agent instance creates a private environment with its own cache, cookies, session data and BrowserEmulator. No data is shared between instances -- each operates within an airtight sandbox to ensure no identities leak across requests.

Default Instance

A default instance is automatically initialized and available as the default export of secret-agent.

The default instance can receive configuration via command line arguments. Any args starting with --input.* will be processed. The resulting json object is available as agent.input

// script.js
const agent = require('secret-agent');

console.log(agent.input); // { secret: "true", agent: "true" }
$ node script.js --input.secret=true --input.agent=true

Constructor

new Agent(options)

Creates a new sandboxed browser instance with unique user session and fingerprints. Or pass in an existing UserProfile to reconstruct a previously used user session.

You can optionally await an instance (or constructor) to cause the connection to the underlying SecretAgent to be initialized. If you don't await, the connection will be established on the first call.

Note: If you provide a name that has already been used to name another instance then a counter will be appended to your string to ensure its uniqueness. However, it's only unique within a single NodeJs process (i.e., rerunning your script will reset the counter).

const { Agent } = require('secret-agent');

(async () => {
  // connection established here
  const agent = await new Agent({
    userAgent: '~ mac 13.1 & chrome > 14'
  });
})();

Arguments:

  • options object Accepts any of the following:
    • connectionToCore options | ConnectionToCore. An object containing IConnectionToCoreOptions used to connect, or an already created ConnectionToCore instance. Defaults to automatically booting up and connecting to a local Core.
    • name string. This is used to generate a unique sessionName.
    • userAgent strong. This sets your browser's user agent string. Prefixing this string with a tilde (~) allows for dynamic options.
    • browserEmulatorId string defaults to default-browser-emulator. Chooses the BrowserEmulator plugin which emulates the properties that help SecretAgent look like a normal browser.
    • humanEmulatorId string defaults to default-human-emulator. Chooses the HumanEmulator plugin which drives the mouse/keyboard movements.
    • geolocation IGeolocation. Overrides the geolocation of the user. Will automatically grant permissions to all origins for geolocation.
      • latitude number. Latitude between -90 and 90.
      • longitude number. Longitude between -180 and 180.
      • accuracy number. Non-negative accuracy value. Defaults to random number 40-50.
    • timezoneId string. Overrides the host timezone. A list of valid ids are available at unicode.org
    • locale string. Overrides the host languages settings (eg, en-US). Locale will affect navigator.language value, Accept-Language request header value as well as number and date formatting rules.
    • viewport IViewport. Sets the emulated screen size, window position in the screen, inner/outer width and height. If not provided, the most popular resolution is used from statcounter.com.
      • width number. The page width in pixels (minimum 0, maximum 10000000).
      • height number. The page height in pixels (minimum 0, maximum 10000000).
      • deviceScaleFactor number defaults to 1. Specify device scale factor (can be thought of as dpr).
      • screenWidth? number. The optional screen width in pixels (minimum 0, maximum 10000000).
      • screenHeight? number. The optional screen height in pixels (minimum 0, maximum 10000000).
      • positionX? number. Optional override browser X position on screen in pixels (minimum 0, maximum 10000000).
      • positionY? number. Optional override browser Y position on screen in pixels (minimum 0, maximum 10000000).
    • blockedResourceTypes BlockedResourceType[]. Controls browser resource loading. Valid options are listed here.
    • userProfile IUserProfile. Previous user's cookies, session, etc.
    • input object. An object containing properties to attach to the agent. NOTE: if using the default agent, this object will be populated with command line variables starting with --input.{json path}. The {json path} will be translated into an object set to agent.input.
    • showReplay boolean. Whether or not to show the Replay UI. Can also be set with an env variable: SA_SHOW_REPLAY=true.
    • upstreamProxyUrl string. A socks5 or http proxy url (and optional auth) to use for all HTTP requests in this session. Dns over Tls requests will also use this proxy, if provided. The optional "auth" should be included in the UserInfo section of the url, eg: http://username:password@proxy.com:80.

Properties

agent.activeTab

Returns a reference to the currently active tab.

Type: Tab

agent.coreHost

The connectionToCore host address to which this Agent has connected. This is useful in scenarios where a Handler is round-robining connections between multiple hosts.

Type: Promise<string>

agent.document
W3C

Returns a reference to the main Document for the active tab.

Type: SuperDocument

Alias for activeTab.document

agent.frameEnvironments

Returns a list of FrameEnvironments loaded for the active tab.

Type: Promise<FrameEnvironment[]>.

agent.input

Contains the input configuration (if any) for this agent. This might come from:

NOTE: if using the default agent, this object will be populated with command line variables starting with --input.*. The parameters will be translated into an object set to agent.input.

Type: Object

agent.lastCommandId

An execution point that refers to a command run on this instance (waitForElement, click, type, etc). Command ids can be passed to select waitFor* methods to indicate a starting point to listen for changes.

Type: Promise<number>

Alias for activeTab.lastCommandId

agent.mainFrameEnvironment

Returns a reference to the document of the mainFrameEnvironment of the active tab.

Alias for tab.mainFrameEnvironment.document.

Type: SuperDocument

agent.meta

Retrieves metadata about the agent configuration:

  • sessionId string. The session identifier.
  • sessionName string. The unique session name that will be visible in Replay.
  • browserEmulatorId string. The id of the Browser Emulator in use.
  • humanEmulatorId string. The id of the Human Emulator in use.
  • timezoneId string. The configured unicode TimezoneId or host default (eg, America/New_York).
  • locale string. The configured locale in use (eg, en-US).
  • geolocation IGeolocation. The configured geolocation of the user (if set).
  • viewport IViewport. The emulated viewport size and location.
  • blockedResourceTypes BlockedResourceType[]. The blocked resource types.
  • upstreamProxyUrl string. The proxy url in use for this agent.
  • userAgentString string. The user agent string used in Http requests and within the DOM.

Type: Promise<IAgentMeta>

agent.output

Agent output is an object used to track any data you collect during your session. Output will be shown in Replay during playback for easy visual playback of data collection.

Output is able to act like an Array or an Object. It will serialize properly in either use-case.

NOTE: any object you assign into Output is "copied" into the Output object. You should not expect further changes to the source object to synchronize.

const agent = require('secret-agent');

(async () => {
  await agent.goto('https://www.google.com');
  const document = agent.document;

  for (const link of await document.querySelectorAll('a')) {
    agent.output.push({ // will display in Replay UI.
      text: await link.textContent,
      href: await link.href,
    });
  }
   
  console.log(agent.output);  
  await agent.close();
})();

Type: Output. An array-like object.

agent.sessionId

An identifier used for storing logs, snapshots, and other assets associated with the current session.

Type: Promise<string>

agent.sessionName

A human-readable identifier of the current Agent session.

You can set this property when calling Handler.dispatchAgent() or Handler.createAgent().

Type: Promise<string>

agent.tabs

Returns all open browser tabs.

Type: Promise<Tab[]>

agent.url

The url of the active tab.

Type: Promise<string>

Alias for Tab.url

agent.Request
W3C

Returns a constructor for a Request object bound to the activeTab. Proxies to tab.Request. These objects can be used to run browser-native tab.fetch requests from the context of the Tab document.

Type: Request

Alias for Tab.Request

Methods

agent.click(mousePosition)

Executes a click interaction. This is a shortcut for agent.interact({ click: mousePosition }). See the Interactions page for more details.

Arguments:

Returns: Promise

agent.close()

Closes the current instance and any open tabs.

Returns: Promise

agent.closeTab(tab)

Close a single Tab. The first opened Tab will become the focused tab.

Arguments:

  • tab Tab The Tab to close.

Returns: Promise<void>

Alias for Tab.close()

agent.configure(options)

Update existing configuration settings.

Arguments:

  • options object Accepts any of the following:
    • userProfile IUserProfile. Previous user's cookies, session, etc.
    • timezoneId string. Overrides the host timezone. A list of valid ids are available at unicode.org
    • locale string. Overrides the host languages settings (eg, en-US). Locale will affect navigator.language value, Accept-Language request header value as well as number and date formatting rules.
    • viewport IViewport. Sets the emulated screen size, window position in the screen, inner/outer width. (See constructor for parameters).
    • blockedResourceTypes BlockedResourceType[]. Controls browser resource loading. Valid options are listed here.
    • upstreamProxyUrl string. A socks5 or http proxy url (and optional auth) to use for all HTTP requests in this session. The optional "auth" should be included in the UserInfo section of the url, eg: http://username:password@proxy.com:80.
    • connectionToCore options | ConnectionToCore. An object containing IConnectionToCoreOptions used to connect, or an already created ConnectionToCore instance. Defaults to booting up and connecting to a local Core.

Returns: Promise

See the Configuration page for more details on options and its defaults. You may also want to explore BrowserEmulators and HumanEmulators.

agent.detach(tab[, key])

Detach the given tab into a "Frozen" state. The FrozenTab contains a replica of the DOM and layout at the moment of detachment, and supports all the readonly activities of a normal Tab (eg, querySelectors, getComputedVisibility, getComputedStyle).

FrozenTabs have a unique attribute in that any queries you run against them will be "learned" on an initial run, and pre-fetched on subsequent runs. This means you can very quickly iterate through all the data you want on a page after you've loaded it into your desired state.

NOTE: you can detach the same Tab multiple times per script. Each instance will contain DOM frozen at the time it was detached.

Arguments:

  • tab Tab. An existing tab loaded to the point you wish to freeze
  • key string. Optional extra identifier to differentiate between runs in a loop. This can be useful if you are looping through a list of links and detaching each Tab but have specific extraction logic for each link. NOTE: if your looping logic is the same, changing this key will decrease performance.

Returns: FrozenTab

await agent.goto('https://chromium.googlesource.com/chromium/src/+refs');
await agent.activeTab.waitForLoad(LocationStatus.DomContentLoaded);

const frozenTab = await agent.detach(agent.activeTab);
const { document } = frozenTab;

const versions = agent.output;
// 1.  First run will run as normal.
// 2+. Next runs will pre-fetch everything run against the frozenTab
// NOTE: Every time your script changes, SecretAgent will re-learn what to pre-fetch.
const wrapperElements = await document.querySelectorAll('.RefList');
for (const elem of wrapperElements) {
  const innerText = await elem.querySelector('.RefList-title').innerText;
  if (innerText === 'Tags') {
    const aElems = await elem.querySelectorAll('ul.RefList-items li a');

    for (const aElem of aElems) {
      const version = await aElem.innerText;
      versions.push(version);
    }
  }
}
await agent.close();

agent.exportUserProfile()

Returns a json representation of the underlying browser state for saving. This can later be restored into a new instance using agent.configure({ userProfile: serialized }). See the UserProfile page for more details.

Returns: Promise<IUserProfile>

agent.focusTab(tab)

Bring a tab to the forefront. This will route all interaction (click, type, etc) methods to the tab provided as an argument.

Arguments:

  • tab Tab The Tab which will become the activeTab.

Returns: Promise<void>

Alias for Tab.focus()

agent.interact(interaction[, interaction, ...])

Executes a series of mouse and keyboard interactions.

Arguments:

Returns: Promise

Refer to the Interactions page for details on how to construct an interaction.

agent.scrollTo(mousePosition)

Executes a scroll interaction. This is a shortcut for agent.interact({ scroll: mousePosition }). See the Interactions page for more details.

Arguments:

  • mousePosition MousePosition

Returns: Promise

agent.type(keyboardInteraction[, keyboardInteraction, ...])

Executes a keyboard interactions. This is a shortcut for agent.interact({ type: string | KeyName[] }).

Arguments:

Returns: Promise

Refer to the Interactions page for details on how to construct keyboard interactions.

agent.use(plugin)

Add a plugin to the current instance. This must be called before any other agent methods.

Arguments:

  • plugin ClientPlugin | array | object | string

Returns: this The same Agent instance (for optional chaining)

If an array is passed, then any client plugins found in the array are registered. If an object, than any client plugins found in the object's values are registered. If a string, it must be a valid npm package name available in the current environment or it must be an absolute path to a file that exports one or more plugins -- Agent will attempt to dynamically require it.

Also, if a string is passed -- regardless of whether it's an npm package or absolute path -- the same will also be registered in Core (however, the same is not true for arrays or objects). For example, you can easily register a Core plugin directly from Client:

import agent from '@secret-agent';

agent.use('@secret-agent/tattle-plugin');

The following three examples all work:

Use an already-imported plugin:

import agent from '@secret-agent';
import ExecuteJsPlugin from '@secret-agent/execute-js-plugin';

agent.use(ExecuteJsPlugin);

Use an NPM package name (if it's publicly available):

import agent from '@secret-agent';

agent.use('@secret-agent/execute-js-plugin');

Use an absolute path to file that exports one or more plugins:

import agent from '@secret-agent';

agent.use(require.resolve('./CustomPlugins'));

agent.waitForNewTab()

Wait for a new tab to be created. This can occur either via a window.open from within the page javascript, or a Link with a target opening in a new tab or window.

Returns: Promise<Tab>

const url = 'https://dataliberationfoundation.org/nopost';
const { document, activeTab } = agent;

await agent.goto('http://example.com');

// ...
// <a id="newTabLink" href="/newPage" target="_blank">Link to new target</a>
// ...

await document.querySelector('#newTabLink').click();
const newTab = await agent.waitForNewTab();

await newTab.waitForPaintingStable();

Aliased Tab Methods

Agent instances have aliases to all top-level Tab methods. They will be routed to the activeTab.

agent.fetch(requestInput, requestInit)
W3C

Alias for Tab.fetch()

agent.getFrameEnvironment(frameElement)

Alias for Tab.getFrameEnvironment()

agent.getComputedStyle(element, pseudoElement)
W3C

Alias for Tab.getComputedStyle()

agent.getJsValue(path)

Alias for Tab.getJsValue()

agent.goBack(timeoutMs?)

Alias for Tab.goBack

agent.goForward(timeoutMs?)

Alias for Tab.goForward

agent.goto(href, timeoutMs?)

Alias for Tab.goto

agent.getComputedVisibility(element)

Alias for Tab.getComputedVisibility

agent.reload(timeoutMs?)

Alias for Tab.reload

agent.takeScreenshot(options?)

Alias for Tab.takeScreenshot

agent.waitForFileChooser(options)

Alias for Tab.waitForFileChooser()

agent.waitForElement(element, options)

Alias for Tab.waitForElement

agent.waitForLocation(trigger, options)

Alias for Tab.waitForLocation

agent.waitForMillis(millis)

Alias for Tab.waitForMillis

agent.waitForPaintingStable()

Alias for Tab.waitForLoad(PaintingStable)

agent.waitForResource(filter, options)

Alias for Tab.waitForResource

Edit this page on GitHub