Monitor API
Added support for monitoring browser windows for specific conditions.
You can monitor browser windows in several ways:
Natural language conditions
Specify conditions in plain English that describe what you’re waiting for:
- “Determine if the user appears to be signed in to the website”
- “Check if the shopping cart contains more than 3 items”
- “Wait for the success message to appear”
Configurable monitoring options
Fine-tune the monitoring behavior with:
intervalSeconds
: How frequently to check the condition (default: 5 seconds)timeoutSeconds
: Maximum time to wait for the condition (default: 30 seconds)includeVisualAnalysis
: Enable visual analysis for enhanced condition detection
Resource management
Control costs and execution time:
costThresholdCredits
: Set credit limits for the monitoring operationtimeThresholdSeconds
: Set maximum execution time limits
When visual analysis is enabled, monitoring operations may use more credits but can detect visual changes more accurately
For more information, see the API reference.
Chrome Extension support
Added support for installing and using Chrome extensions in browser sessions.
For more information, see the Chrome Extension support guide.
Scroll interaction
Added support for scrolling in browser windows. Scroll
You can now programmatically scroll in three ways:
Scroll by relative amount
Use scrollBy
to scroll a specific amount horizontally or vertically from the current position:
xAxis
: Scroll horizontally by pixels or percentage (e.g. “100px” or “50%”). Positive values scroll right, negative values scroll left.yAxis
: Scroll vertically by pixels or percentage (e.g. “100px” or “50%”). Positive values scroll down, negative values scroll up.
Scroll to page edges
Use scrollToEdge
to quickly navigate to the edges of the page:
xAxis
: Scroll to “left” or “right” edgeyAxis
: Scroll to “top” or “bottom” edge
Scroll to specific elements
Use scrollToElement
to scroll to elements matching a natural language description (e.g. “the search box”, “the submit button at the bottom of the form”).
scrollToElement
is provided, it takes precedence over scrollBy
and scrollToEdge
configurations, and scrollBy
/scrollToEdge
will be ignoredFor more details, see the API reference.
Improved Residential Proxies
Sticky sessions
Added support for sticky sessions so you can maintain the same proxy IP address for up to 30 minutes at a time.
Country support
You can now specify a country for your proxy.
Better defaults
If you are already using Airtop’s integrated proxy, all future requests will automatically get a proxy in the 🇺🇸 USA with sticky sessions.
If you wish for the previous behavior, you can use our new configuration options: change the country to global
and sticky to false
.
For more information, see our proxy documentation.
Visual Analysis mode Result Selection Strategy
Visual analysis now supports a resultSelectionStrategy
property that controls how elements are selected when multiple potential matches are found. The available strategies are:
auto
(default) - Let the Airtop decide the best strategy.first
- Select the first matching element foundbestMatch
- Select the element with the highest confidence score. Additional LLM credits will be used on this strategy.
This gives you more control over which element is selected when there are multiple potential matches on the page.
Check out the visual analysis config for more details.
Screenshots
You can now take screenshots of the browser window! Check out the screenshot reference for more details.
Visual Analysis mode Scan
Visual analysis now supports a new scan
scope mode that progressively scrolls through the page while taking screenshots.
This mode is useful for sites that have compatibility or accuracy issues with the default page
mode.
You can configure the maximum number of scrolls and delay between scrolls using the maxScanScrolls
and scanScrollDelay
parameters in the visualAnalysis
config.
Check out the visual analysis config for more details.
Enhanced Prompts
The Playground in our Studio allows you to automatically enhance your prompts and generate a JSON schema for the output. Detailed prompts and schemas help the model produce much more accurate results and this feature makes it much easier to create them.
Right Click + Double Click
Click interactions now support right click and double click. Simply add the rightClick
or doubleClick
parameter to the request. More details in the interactions guide.
Name based profiles
We’ve updated our profile management to now use names instead of IDs. We deprecated the profileID
based API methods and replaced them with profileName
based methods. This will make it easier to manage profiles by creating and saving profiles under the same name.
Code Execution in the Studio
Added support for running Typescript code in the Studio, as well as sharing the code snippets with other users. Now you can get started quickly and share a fully functional script without having to open an IDE. Check it out in the Studio.
Make Integration
We launched a new integration with Make. You can now use Airtop with Make to access content from the web and interact with sites. For more information, see the guide documentation for Make.
Batch Operations
Added support for batch operations to the Node and Python SDKs. You can now use the batch operate helpers to repeatedly run an operation on a list of URLs. Concurrency and cleanup will be handled automatically. For more information, see the guide documentation for batch operations.