Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Brower Automation

    1

    Drive a headed Chrome browser through 41 MCP tools — observe, click, fill, scrape, screenshot — with verification-first, closed-loop automation.

    $2

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Brower Automation

    Brower Automation

    Example session with this skill installed

    Open https://example.com, observe the page, dismiss any cookie banner, click the primary call-to-action, and confirm the destination page rendered.

    • Read your context and instructions
    • Compiled the brower automation

    Summary

    Opened example.com, dismissed the cookie banner, clicked the primary call-to-action, and verified the destination page rendered. No guessed URLs — every target came from an observed ref.

    Steps taken

    1. session.create → headed Chrome, viewport pinned 1280×720.
    2. page.open { url: "https://example.com" } → _ready: true, lifecycle resolved.
    3. page.handleOverlays { close_cookie_banner: true } → detected_overlay: false.
    4. page.observe { detail: "standard" } → minted refs e1…e23; primary CTA e7 (role=button, name="Get started").
    5. page.click { target:{ref:"e7"} } → clicked:true, verified:true, verification:"rendered_destination", resolution_tier:1.
    6. page.observe { detail:"minimal" } → confirmed destination heading present.
    MetricBeforeAfter
    Clicks verified0%100%
    Guessed URLs10
    Resolution tier-1

    Next steps

    • Fill the application form field-by-field (one field at a time for controlled inputs), then page.press Tab to commit.
    • If a CAPTCHA appears, session.pause and let a human solve it, then session.resume.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Browser Automation

    A global skill that turns any compatible AI agent into a reliable, verification-first browser operator. Instead of guessing selectors and hoping clicks land, your agent drives a real headed Chrome through 41 semantic MCP tools and only proceeds when an action is verified.

    What it solves

    • Guess-and-stall loops — every action returns post-action state and a verified flag, so the agent never assumes a screen it hasn't observed.
    • Fragile selectors — refs minted by page.observe resolve to the exact node (resolution tier 1) and survive minor DOM churn.
    • Silent failures — clicked: true is never treated as success; verified: true is required before proceeding.
    • Overlay/cookie traps — built-in overlay discovery and dismissal with safe-intent checks.

    Key capabilities

    • Navigation, tabs, scoped observation (minimal/standard/full), instant page.probe.
    • Vision: real pixel screenshots (viewport, full-page, element crops) with device/CSS pixel conversion.
    • Interaction: click, drag (trusted input with real timing), fill with recovery ladder, press, select, check, scroll, paginate, viewport pinning, real file upload.
    • JavaScript in main frame and iframes, text/experience extraction, network capture.
    • Human-in-the-loop: headed browser, pause/resume for login, CAPTCHA, 2FA; persistent profile keeps cookies and logins.
    • Bot-evasion stealth script; realtime status banner (controlling/thinking/awaiting).

    Why buy

    You get a scenario-tested runtime contract — rules, closed loop, tool catalog, error taxonomy, timing fields, and recovery recipes — so your agent stops clicking strange "Apply" buttons and starts shipping verified browser workflows. Works across Claude Code, Cursor, Codex CLI, and 18+ compatible agents.

    User may have to prompt their agent to install it in their agent's configuration file. Google-account sign-in is blocked on Puppeteer-launched Chrome (use cdp_url attach or a copied profile). Headless mode is not supported. Refs retire after the 6000 most recent per session. Some apps commit on blur, not on input. Consecutive fills on controlled inputs can race — fill one field at a time.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 28 days ago

    Requires

    Chrome or Chromium
    Node.js 18+

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Browser
    Network Access

    File Scopes

    browser-profile
    dump
    *.md
    Listed28 days ago
    Updated12 days ago

    What's inside

    Frequently Asked Questions

    Tool auto-discovers it; or add the folder path to skills in config.json.

    Load the skill, then ask your agent to browse, click, fill forms, scrape, or screenshot. It observes first, acts on refs, and verifies every click before proceeding.

    Yes via session.pause/session.resume for human-in-the-loop, with a persistent profile keeping cookies and logins. Google sign-in on a launched Chrome is blocked by Google — attach via cdp_url or use a copied profile.

    New in Agents & Orchestration