🎭Connect Playwright, Puppeteer or Selenium

The pattern is the same in every language: ask LoginDeck to open the profile, then attach to the endpoint it gives you. Never launch the browser yourself — a browser your library started is a browser with none of the profile's fingerprint, proxy or cookies, and every detector knows what it looks like.

First turn the API on and copy your token: Get started with the local API.

🔑 What browser/start hands back#

curl -s http://127.0.0.1:19222/api/v1/browser/start \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"name": "My profile"}'

The data object contains the endpoint, said three ways because the libraries want it different ways:

Also in there: proxy (without its password), timezone, exit (the IP, country, city and ISP the session actually comes out of) and headless.

🎭 Playwright (JavaScript)#

import { chromium } from 'playwright';

const AO = { base: 'http://127.0.0.1:19222', token: 'YOUR_TOKEN' };

async function ao(path, body) {
  const res = await fetch(AO.base + path, {
    method: body ? 'POST' : 'GET',
    headers: {
      Authorization: 'Bearer ' + AO.token,
      ...(body ? { 'Content-Type': 'application/json' } : {}),
    },
    body: body ? JSON.stringify(body) : undefined,
  });
  const { ok, data, error } = await res.json();
  if (!ok) throw new Error(error);
  return data;
}

const session = await ao('/api/v1/browser/start', { name: 'My profile' });

const browser = await chromium.connectOverCDP(session.ws);

// The profile's OWN context. newContext() would give you a blank one
// with none of the profile's cookies.
const ctx = browser.contexts()[0];
const page = ctx.pages()[0] || await ctx.newPage();

await page.goto('https://iphey.com');
console.log(await page.title());

// Leave the browser to LoginDeck, and close it through the API.
await ao('/api/v1/browser/stop', { name: 'My profile' });

🐕 Puppeteer (JavaScript)#

import puppeteer from 'puppeteer-core';

// ao() as above
const session = await ao('/api/v1/browser/start', { name: 'My profile' });

const browser = await puppeteer.connect({
  browserWSEndpoint: session.ws,
  // null keeps the profile's own window size
  defaultViewport: null,
});

const [page] = await browser.pages();
await page.goto('https://iphey.com');
console.log(await page.title());

// disconnect leaves the browser open — close it through the API
browser.disconnect();
await ao('/api/v1/browser/stop', { name: 'My profile' });

Use puppeteer-core: you are attaching to LoginDeck's browser, so there is no reason to download Puppeteer's own Chromium.

🐍 Playwright (Python)#

import requests
from playwright.sync_api import sync_playwright

AO = "http://127.0.0.1:19222"
H = {"Authorization": "Bearer YOUR_TOKEN"}

s = requests.post(
    f"{AO}/api/v1/browser/start",
    headers=H, json={"name": "My profile"},
).json()["data"]

print(s["ws"], s["exit"])

with sync_playwright() as p:
    browser = p.chromium.connect_over_cdp(s["ws"])
    ctx = browser.contexts[0]
    page = ctx.pages[0] if ctx.pages else ctx.new_page()
    page.goto("https://iphey.com")
    print(page.title())

requests.post(f"{AO}/api/v1/browser/stop", headers=H, json={"name": "My profile"})

🧪 Selenium (Python)#

Selenium attaches to the same debugging port through debugger_address. It needs a chromedriver that matches the engine's major version, which engine.version in the reply tells you (Chrome/155.0.0.0 → chromedriver 155).

import requests
from selenium import webdriver
from selenium.webdriver.chrome.options import Options

AO = "http://127.0.0.1:19222"
H = {"Authorization": "Bearer YOUR_TOKEN"}

s = requests.post(
    f"{AO}/api/v1/browser/start",
    headers=H, json={"name": "My profile"},
).json()["data"]

print("engine:", s["engine"]["version"])   # which chromedriver you need

opts = Options()
opts.debugger_address = f"127.0.0.1:{s['debug_port']}"

driver = webdriver.Chrome(options=opts)    # attaches, does not launch
driver.get("https://iphey.com")
print(driver.title)

# Do NOT call driver.quit() — it would kill the browser out from under
# LoginDeck. Detach, then stop the session through the API.
requests.post(f"{AO}/api/v1/browser/stop", headers=H, json={"name": "My profile"})

In Java or C#, the same idea: ChromeOptions.setExperimentalOption("debuggerAddress", "127.0.0.1:" + debugPort).

📋 Rules that save you an afternoon#

• Do not launch, attach. No chromium.launch(), no puppeteer.launch(), no plain webdriver.Chrome() without debugger_address.
• Stop the session through the API, not by closing the browser from your library. That keeps the app's state, the profile's lease and the cookie save correct.
• One session per profile. A second browser/start on a profile that is already open returns the session that is already running; a profile opened by hand from the app cannot be taken over (this profile is already running without an automation port — stop it first).
• Windowed unless you have a reason. {"headless": true} is allowed and the reply warns you: no real display, a software GPU path, and several APIs behave differently.
• Chromium only. Firefox profiles cannot be automated yet.
• Give the app a moment. browser/start returns once the browser has a debugging port, so by the time you have the ws URL you can connect straight away — no sleep needed.

🧠 Or let a Flow do the clicking#

If what you are writing is "open page, click, type, read", that already exists as a Flow, and a Flow survives the site being redesigned better than a selector-based script does. Your script can start one and wait for the outcome:

curl -s http://127.0.0.1:19222/api/v1/flows/run \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"flow": "My flow", "name": "My profile", "wait": true}'

See What Flows are and the Local API reference.

ℹ️ Good to know#

• The endpoint is on 127.0.0.1 and changes every launch — read it from the reply each time rather than hard-coding a port.
• GET /api/v1/browser/list gives you every running session and its endpoint, so a script that lost its connection can reattach.
• LoginDeck must stay open. It is the app that owns the browsers.
• If engine.patched comes back false, the session is on stock Chrome and the canvas, WebGL, audio and font masking is inert. See What no antidetect browser can change.