PrompteryxBeta
    ⌘K
    PlatformSupport
    Menu
    IntroductionInstall Desktop & extensionQuick StartPlatform Overview
    OverviewWhat you can makeAI Browser Agent in the chatModels and AutoConnectorsSchedulesProjectsExplore and publishingShare linksMemory and SkillsPersonal agent (beta)Welcome, settings and skinsPlans, credits and top-upsPrompteryx DesktopGroups (pilot)Phone app (early access)
    OverviewCreating WorkflowsTesting & DebuggingScheduling
    OverviewGetting StartedCommandsRecordingSettingsStructured OutputParallel AgentsTips & Best Practices
    OverviewInstallationSidebar AgentUsing Attached FilesExecute on Current TabExporting to Visual Studio
    Overview
    OverviewOpen PageClickClick CoordinatesFillExtract TextScreenshotWaitPress KeyFocus ElementBlur ElementSelect TextHighlight ElementSet ViewportGet Bounding BoxScrollHoverSelect OptionAssertionsCookiesEvaluate JavaScriptExample Workflows
    OverviewPrompteryx Browser AgentOpenAI
    OverviewSchedule TriggerWebhook Trigger
    OverviewGoogle SheetsGoogle CalendarConnect Hub (3,000+ apps)Example Workflows
    AI Vision SettingsChrome Automation ProfilesCloud BrowserBrowser Profiles & Persistent ContextVideo RecordingExecution LogsSharing & CollaborationEdit History
    Selectors GuideVariables
    Start hereGet an API keyRun the AI browser agentDrive a cloud browserRun workflows from anywhereUse connected apps (Connect Hub)SDK (TypeScript)CLIMCP for AI assistants
    REST APIWebhooksMCP toolsCloud Browser APICloud Browser quickstartMigrate from Browserbasen8nZapierMake.com

    Need help? Contact Support

    1. Documentation
    3. AI Browser
    5. Commands

    Commands & Prompting

    Prompting Guide

    Learn how to write effective prompts for the AI Browser Agent

    The AI Browser Agent understands natural language instructions. Unlike traditional automation that requires precise selectors and scripts, you simply describe what you want in plain English. This guide will help you write prompts that get the best results.

    The AI Browser Agent composer with a complete command typed in: go to a site, search, apply a filter, then extract specific fields from the first results
    A well-formed command in the composer — the site, the search, the filter and exactly what to extract, all in one instruction.

    Writing Effective Prompts

    Be Specific About the Target
    Good
    • "Go to amazon.com and search for headphones"
    • "Navigate to linkedin.com/jobs and find software engineer roles"
    • "Open the Google Sheets document at [URL]"
    Avoid
    • "Search for headphones" (which site?)
    • "Find jobs" (too vague)
    • "Open my document" (which one?)
    Describe Actions Clearly
    Good
    • "Click the 'Sign In' button in the top right corner"
    • "Type 'john@example.com' in the email field"
    • "Select 'California' from the state dropdown"
    • "Scroll down until you see the pricing section"
    Avoid
    • "Login" (how? where?)
    • "Enter my email" (what email?)
    • "Pick a state" (which one?)
    • "Go down" (how far?)
    Include Context for Multi-Step Tasks

    For complex tasks, break them down or describe the full workflow upfront:

    Multi-step example:

    "Go to linkedin.com, search for 'product manager' jobs in 'San Francisco', filter by 'Past 24 hours', then extract the job title, company name, and link for the first 5 results."

    Or use follow-up messages to guide step by step:

    You: "Go to linkedin.com/jobs"

    Agent: Done, navigated to LinkedIn Jobs

    You: "Search for 'product manager' in 'San Francisco'"

    Agent: Done, showing results

    You: "Extract the first 5 job titles"

    Common Command Patterns

    Navigation
    • "Go to [URL]"
    • "Navigate to [website]"
    • "Open [website] in a new tab"
    • "Go back to the previous page"
    • "Refresh the page"
    Clicking
    • "Click the [button/link/element]"
    • "Click on '[visible text]'"
    • "Click the button that says '[text]'"
    • "Right-click on [element]"
    • "Double-click [element]"
    Typing & Forms
    • "Type '[text]' in the [field]"
    • "Enter '[value]' in the [input]"
    • "Fill the form with [values]"
    • "Select '[option]' from the dropdown"
    • "Check the '[checkbox label]' checkbox"
    • "Clear the [field]"
    Data Extraction
    • "Extract [data] from the page"
    • "Get the text of [element]"
    • "Copy all the [items] into a list"
    • "Take a screenshot"
    • "What does the page say about [topic]?"
    Scrolling & Waiting
    • "Scroll down"
    • "Scroll to the bottom of the page"
    • "Scroll until you see [element]"
    • "Wait for the page to load"
    • "Wait for [element] to appear"
    Search
    • "Search for '[query]'"
    • "Find '[text]' on this page"
    • "Look for products matching '[criteria]'"
    • "Search Google for '[query]'"

    Advanced Tips

    Use Visual Descriptions
    The agent uses computer vision to "see" the page. You can reference elements by their appearance:
    • "Click the blue button at the top"
    • "Select the red 'Delete' icon next to the item"
    • "Click the profile picture in the top right corner"
    Provide Feedback
    If the agent makes a mistake, tell it:
    • "That's not right, click the other button instead"
    • "Go back and try clicking on the image instead"
    • "The form is wrong, clear it and start again"
    Handle Dynamic Content
    For pages with loading spinners or dynamic content:
    • "Wait for the results to load, then extract the data"
    • "After the spinner disappears, click the first item"
    • "Keep scrolling until all products are loaded"

    Limitations

    While powerful, the AI Browser Agent has some limitations:

    • Can be stopped by CAPTCHAs or bot detection (Settings → Cloud Browser has the Auto-solve CAPTCHAs and residential proxy options)
    • Cannot interact with native OS dialogs (file upload dialogs, print dialogs)
    • May struggle with very complex or heavily animated pages
    • Cannot access content behind passwords it doesn't know