// DOCUMENTATION

Playwright migration

API mapping from Playwright to Koko SDK and known gaps.

SDK Reference

The Koko SDK public API is modeled after Playwright. Most scripts change only the import and connection line.

Connection

Playwright	Koko SDK
`chromium.connectOverCDP(url)`	`Browser.connect(url)`
`chromium.launch()`	`Browser.launch({ profile })`

Core API map

Playwright	Koko SDK	Notes
`browser.newContext()`	`browser.newContext()`	Client-side grouping
`context.newPage()`	`context.newPage()`
`page.goto()`	`page.goto()`	+ `waitUntil: "done"`
`page.locator()`	`page.locator()`	CSS selector
`page.getByRole()`	`page.getByRole()`	DOM heuristics
`locator.click()`	`locator.click()`
`page.evaluate()`	`page.evaluate()`
`page.screenshot()`	`page.screenshot()`	PNG/JPEG via CDP
`page.pdf()`	`page.pdf()`
`page.waitForSelector()`	`page.waitForSelector()`	Uses performSearch

Koko-only additions

API	Description
`page.markdown()`	Token-efficient page text
`page.semanticTree()`	Pruned a11y DOM
`page.findElement()`	Role/name → backendNodeId
`page.node()`	NodeHandle actions
`page.searchGoogle()`	SERP extract + block detect
`page.armDialog()`	Pre-arm JS dialogs
`page.extract()`	Crawler-optimized extract
`captureSessionState()`	Cookies + storage snapshot

Not in SDK (use CDP or MCP)

expect() assertions — use Vitest, Node assert, etc.
page.route() / request interception
frameLocator() / iframe switching
keyboard / mouse standalone objects
browser.newBrowserCDPSession() tracing APIs
Full getByRole accessibility tree parity

Migration checklist

Replace import { chromium } with import { Browser }
Start Koko CDP server or use Browser.launch()
Swap connectOverCDP → Browser.connect
Optionally adopt waitUntil: "done" for stricter navigation
Replace brittle selectors with NodeHandle for agent flows