ceven

Browser Tool

Controls a headless browser to scrape dynamic websites, fill out complex web forms, and execute multi step navigation tasks on your behalf.

Try Browser Tool in Ceven

Ask Ceven anything
Standard

Why use Ceven?

  1. AI native Browser Tool integration

    • Describe the outcome and Ceven picks the right Browser Tool calls, fills the parameters, and checks the result.
    • Structured, agent friendly tool schemas so each call runs reliably instead of by guesswork.
    • Rich coverage for reading, writing, and querying your Browser Tool data, across all 5 of its actions.

  2. Managed auth

    • Built in OAuth with automatic token refresh and rotation.
    • One place to manage, scope, and revoke Browser Tool access.
    • Per user and per environment credentials instead of shared keys.

  3. Agent optimized design

    • Actions are tuned from real success and error rates so reliability climbs over time.
    • Full execution logs so you always know what ran in Browser Tool, when, and on whose behalf.
    • The agent pauses and asks when Browser Tool is unclear instead of plowing ahead.

  4. Enterprise grade security

    • Fine grained access so you control which agents and people can reach Browser Tool.
    • Least privilege by default, read scopes first and only the writes a workflow needs.
    • A full audit trail of every Browser Tool action to support review and sign off.

Supported tools

Every action Ceven's agents can run on Browser Tool, and when to use it.

Navigate to URL
Use this to start a session and load a specific web page. This is the required first step for every browser workflow.
Fetch Webpage Content
Pull the current HTML content of the page to understand the state and find elements for future actions.
Mouse Click
Perform a precision click at specific coordinates. Use this for buttons or links that lack standard identifiers.
Type Text
Insert text into a focused input field using human like typing speeds to avoid bot detection.
Scroll Page
Move the page view up or down. Use this when elements are not visible in the initial page fetch.
Take Screenshot
Capture a visual image of the current viewport to verify a page state or document a visual error.
AI Perform Web Task
Delegate a complex multi step goal to the AI. Use this only for unpredictable flows with ten or more steps.
Paste Text
Insert content from the clipboard into the current cursor position for fast form filling.
Get Clipboard Content
Read the current text stored in the system clipboard for data transfer between pages.
Set Clipboard Content
Store specific text in the clipboard to prepare it for a later paste action.
Keyboard Shortcut
Execute key combinations like Ctrl C or Ctrl V to trigger browser or application commands.
Mouse Move
Position the cursor at specific coordinates to trigger hover effects or reveal hidden tooltips.
Drag and Drop
Move an element from one location to another. Use this for file uploads or reordering lists.
Copy Selected Text
Extract the currently highlighted text on the page and save it to the clipboard.
Mouse Double Click
Execute a double click at coordinates to open files or select specific text blocks.
Mouse Down (Press and Hold)
Press and hold mouse button at coordinates - use for starting custom drag operations, text selections, or long press interactions. must be followed by mouseup action to complete.
Mouse Up (Release Button)
Release mouse button at coordinates - completes drag operations, text selections, and long press interactions. should be used after mousedown to finish mouse button sequences.
Screenshot Webpage
Capture high quality screenshot of any webpage with extensive customization options - perfect for archiving, visual documentation, full page captures, and cross device viewport testing.

18 actions · scroll to see them all

Frequently asked questions

Ceven utilizes the Browser Tool capabilities to mimic human behavior as closely as possible. This includes using a realistic user agent string and implementing variable delays between typing and clicking actions. Instead of sending raw HTTP requests which are easily flagged, the agent interacts with a full browser instance that renders CSS and executes Javascript. For sites with aggressive bot detection, we recommend using the Type Text action with a human like delay set to 100 milliseconds. This reduces the likelihood of triggering security challenges by avoiding the instantaneous input patterns typically associated with automated scripts and basic scrapers.
Yes. The agent can navigate through a sequence of pages by combining the Navigate to URL action with Mouse Click and Scroll Page actions. The agent maintains the session state across these calls, meaning if it logs into a portal on the first page, it remains logged in as it moves to subsequent pages. The typical workflow involves fetching the page content, identifying the link for the next step, clicking that link, and then verifying the new page has loaded by taking a screenshot or fetching the HTML again to confirm the URL has changed as expected.
Because Ceven uses an AI agent to interpret the page content via the Fetch Webpage Content action, it is more resilient than traditional scrapers. Instead of relying on a hard coded path, the agent looks at the current HTML and searches for semantic markers like button text or ARIA labels. If a button moves from the left to the right side of the page, the agent will still find it by its label. If the layout changes so drastically that the element is gone, the agent will recognize the failure and can be programmed to notify you or try an alternative navigation path.
The primary limitation is the rate limit imposed by the target website rather than Ceven itself. If you trigger too many requests in a short window, the target server may issue a 429 Too Many Requests error or present a CAPTCHA. One specific quirk of the Browser Tool is that it operates on a session based model. If you attempt to open too many concurrent browser sessions across different workflows, you may hit a concurrency limit based on your Composio tier. We recommend sequencing your scraping tasks or adding a sleep timer between large batch runs to avoid being flagged as a denial of service attack.
The Browser Tool itself does not include a built in CAPTCHA solver. When the agent encounters a CAPTCHA, it will typically identify it in the page content and signal a failure in the workflow. To handle this, you can set up a human in the loop step where Ceven sends you a screenshot of the CAPTCHA and asks for the solution. Once you provide the text, the agent uses the Type Text action to enter the code and proceed. This ensures that the workflow can recover from security challenges without requiring the agent to possess prohibited cracking capabilities.
Yes. The agent can handle authentication by navigating to a login page, typing the credentials into the username and password fields, and clicking the submit button. Once the session cookie is set in the browser instance, all subsequent actions in that session are authenticated. For higher security environments, you can manually log in and provide the session cookies to the agent, although the standard flow of automating the login form is the most common approach. This allows the agent to access private dashboards and internal company tools that require a secure login.
The agent follows a specific loop of observation and action. First, it calls Fetch Webpage Content to get the HTML. It then analyzes the DOM to find the element that matches the goal, such as a button labeled Submit. If the element has a clear ID or class, the agent uses that. If not, it estimates the screen coordinates based on the element position in the rendered viewport. After performing a Mouse Click, the agent always fetches the page content again or takes a screenshot to verify that the click actually triggered the intended state change before moving to the next step.
Absolutely. Many users employ Ceven for regression testing by creating a workflow that mimics a user journey. For example, the agent can navigate to a staging site, add an item to a cart, and verify that the checkout page loads correctly. By using the Take Screenshot action at each critical step, you create a visual audit trail of the test. If the agent finds that a button is missing or a page fails to load, it can automatically log a ticket in Jira with the screenshot attached, making it a powerful tool for quality assurance teams.

Alternatives to Browser Tool

Other tools that solve a similar problem. Ceven supports these too, so you can switch or run more than one at once.

browserless logobrowserlessApify logoApifySelenium logoSeleniumPlaywright logoPlaywright

Try Ceven on your stack

Plug Ceven on top of the tools you already run. Connect Browser Tool and the rest of your stack, describe the outcome, and its agents handle the work end to end, days of it in minutes.

Get started for free