Error desk

CDP attach failures

connectOverCDP timing out is almost never “the fingerprint.” It’s the Launcher, the port, or the token.

Checklist

1. Multilogin app running on the same machine that calls launcher.mlx.yt:45001.
2. Start with automation_type=playwright (or puppeteer). Without it you may not get an automation port — guide.
3. Use http://127.0.0.1:{data.port} from the start response — not a guessed 9222.
4. Token still valid (~30 minutes per Multilogin help) — token desk.
5. Don’t chromium.launch() a naked browser and expect Multilogin’s profile storage.
6. Mimic cores for Playwright; Stealthfox + Playwright is not available per Multilogin docs.

Symptoms that look like “CDP is broken”

• ECONNREFUSED 127.0.0.1:PORT after agent restart — old port from a previous start response; re-call Launcher start and use the new data.port.
• Connect works once, then flakes — token ~30 min expiry cascading into “port closed”; re-auth first (token desk).
• Attaching to localhost from a remote runner — Launcher binds on the machine running Multilogin; tunnel or run the script there.
• Stealthfox profile + Playwright — Multilogin’s docs still gate Mimic for Playwright; wrong core, not a flaky CDP library.

Sanity connect (after you have a port)

import { chromium } from "playwright";
const browser = await chromium.connectOverCDP(`http://127.0.0.1:${port}`, { timeout: 10_000 });
const context = browser.contexts()[0];
const page = context.pages()[0] || await context.newPage();

Full flow: Playwright guide. Puppeteer twin: Puppeteer guide.

Ads desks

If CDP works but every site smells datacenter, switch to the proxy desk. If auth 401s cascade, stop rewriting Playwright — fix PAM first via API auth.

SAAS50 / MIN50 API map