Chrome API: Change one line and your script runs on our Chrome.

The Chrome API gives you cloud browsers that connect to your existing Puppeteer, Playwright, or CDP code. One long-lived WebSocket per session. No Chromium to install. No servers to manage.

Get API key

Chrome API

1import { chromium } from "playwright";2 3const browser = await chromium.connectOverCDP(4  `wss://cdp-lb.axiom.ai/?token=${process.env.AXIOM_API_KEY}`5);6 7try {8  const context = browser.contexts()[0];9  const page = context.pages()[0] ?? await context.newPage();10  await page.goto("https://example.com");11  await page.screenshot({ path: "out.png", fullPage: true });12} finally {13  await browser.close();14}

Take a screenshot.Read the Chrome API docs

Connect

Your scripts, our browsers.

Instead of launching Chrome locally, your script connects to one of ours over a WebSocket. The rest of the script stays the same.

One line connects

Swap launch() for connect() pointed at our WebSocket. Nothing else changes.

const browser = await chromium.connectOverCDP(
  `wss://cdp-lb.axiom.ai/?token=${process.env.AXIOM_API_KEY}`
);

Any CDP client

Puppeteer, Playwright's connectOverCDP, Python, Go, or raw WebSocket frames.

Full protocol access

Raw CDP, not a wrapper SDK. Every click, wait, and selector stays yours.

No lock-in

Point your script back at launch() and it runs locally again.

In practice

Three things worth trying.

The endpoint is the same in all three. Point Playwright at it with connectOverCDP, point Puppeteer at it with connect, or call a Step API function on the browser you connected to and delete the loop you were about to write.

Playwright over CDP

Playwright, unchanged

connectOverCDP takes the same endpoint, and locators, tracing, and assertions all keep working.

Explore Playwright

const browser = await chromium.connectOverCDP(
  `wss://cdp-lb.axiom.ai/?token=${KEY}`
);

const ctx = browser.contexts()[0];
const page = ctx.pages()[0] ?? await ctx.newPage();

await page.goto("https://example.com");
await page.getByRole("button",
  { name: "Accept" }).click();

Puppeteer over CDP

Puppeteer, connected

puppeteer.connect points at the same WebSocket, so your selectors, waits, and evaluate calls keep working.

Explore Puppeteer

const browser = await puppeteer.connect({
  browserWSEndpoint:
    `wss://cdp-lb.axiom.ai/?token=${KEY}`
});

const [page] = await browser.pages();
await page.goto("https://example.com");
await page.click("button.accept");

CDP plus a step

One call instead of a dozen lines

Step API functions run on the browser you connected to. scrape takes the selector for the data, the next-page button, and how many rows you want, then does the waiting and paging itself. Twenty of them cover the jobs you would otherwise write out, captchas and date pickers included.

Explore the Step API

// your own interception, on your terms
await page.setRequestInterception(true);
page.on("request", r =>
  r.url().endsWith(".png") ? r.abort() : r.continue());

await page.goto("https://example.com/list");

// one call instead of a paging loop
const rows = await axiom.scrape(
  null, ".row", ".next", 500, {});