kernel-typescript-sdk_skill

Overview

This skill lets you build browser automation scripts using the Kernel TypeScript SDK with Playwright, CDP, and remote browser management. It enables server-side execution inside remote browser VMs, programmatic session and profile management, and scalable orchestration for scraping, testing, and deployed actions. Use it to run reliable, production-ready browser automation without local browser dependencies.

How this skill works

Initialize the SDK with a Kernel API key and call resource methods like kernel.browsers.playwright.execute() to run Playwright code inside the remote browser VM. For client-side workflows, obtain a CDP WebSocket URL (browser.cdp_ws_url) and connect local Playwright/Puppeteer to the remote browser. Use dedicated kernel APIs for binary data (screenshots, files) because playwright.execute returns undefined for binaries in TypeScript.

When to use it

• Build server-side Playwright automation that runs inside a remote browser VM
• Manage browser sessions, profiles, proxies, and extensions programmatically
• Scale scraping or testing with browser pools and persistent profiles
• Deploy automation scripts as Kernel actions callable by API
• Use CDP connections for local debugging or complex development workflows

Best practices

• Prefer server-side execution (kernel.browsers.playwright.execute) for production and parallel runs
• Always use return inside executed Playwright code to get response.result back
• Handle binaries with dedicated APIs (computer.captureScreenshot, filesystem.readFile) instead of returning them from execute
• Use snake_case attribute access (e.g., browser.session_id, browser.cdp_ws_url) to interact with resources
• Pre-warm browser pools and reuse profiles for high-volume automation to reduce startup latency

Example use cases

• Headless web scraping at scale with managed browser pools and rotating profiles
• Automated end-to-end testing executed in remote browser VMs without local dependencies
• Deployable actions that perform sign-in flows using persistent profiles and credential providers
• Remote debugging by connecting local Playwright to a browser via CDP WebSocket URL
• Capture screenshots or download files from the VM using kernel.browsers.computer and filesystem APIs