Connect URL

Connecting to a cdpfleet connect URL launches the browser and connects to it in one step, with no launch request first. Use it wherever a tool takes a WebSocket URL: Playwright's connect(), Playwright Test's wsEndpoint, connectOverCDP(), Puppeteer, and AI-agent frameworks that accept a CDP URL. The launch request keeps working as before.

The URL

wss://router01.cdpfleet.com/connect/<browser>?<options>        # Playwright protocol (client 1.60; Rebrowser 1.52)
wss://router01.cdpfleet.com/connect/<browser>/cdp?<options>    # CDP — Chrome, Edge, Brave, Opera, Yandex, Chromium

• API key: the x-api-key header (Playwright and Puppeteer send headers on connect), or ?api_key= for tools that can't send headers.
• Options: the launch options as query parameters — headless=new, channel=beta, profile=shop-login, record=false, proxy=cdpfleet-resi-country-us… Anything that isn't a plain string or true/false (objects, lists, proxy_rules, Camoufox config) goes as one launch= parameter: the JSON launch body, base64url-encoded — or in an x-cdpfleet-launch header as plain JSON. The code builder's Connect URL style writes the URL for you.
• Proxies in URLs: URLs end up in logs, so prefer tokens without passwords — cdpfleet-resi…, cdpfleet-plan…, a saved VPN name — or leave proxy out for the free Community proxy.

Examples

// Playwright (Node.js)
const browser = await chromium.connect('wss://router01.cdpfleet.com/connect/chrome?proxy=cdpfleet-resi-country-us&headless=new', { headers: { 'x-api-key': process.env.CDPFLEET_API_KEY } });

// Playwright Test — playwright.config.ts: every worker gets its own browser
use: { connectOptions: { wsEndpoint: 'wss://router01.cdpfleet.com/connect/chrome?proxy=cdpfleet-resi-country-us&headless=new', headers: { 'x-api-key': process.env.CDPFLEET_API_KEY } } },

// CDP: connectOverCDP (any Playwright version) or Puppeteer
const browser = await chromium.connectOverCDP('wss://router01.cdpfleet.com/connect/chrome/cdp?headless=new', { headers: { 'x-api-key': process.env.CDPFLEET_API_KEY } });
const browser = await puppeteer.connect({ browserWSEndpoint: 'wss://router01.cdpfleet.com/connect/chrome/cdp?headless=new', headers: { 'x-api-key': process.env.CDPFLEET_API_KEY } });

Python: p.chromium.connect(url, headers={"x-api-key": key}). Java, C# and Go connect the same way with their own headers option — the code builder writes all five.

What happens

• The connection runs the same launch as POST /<browser>/session: same options, limits, billing and errors. The session starts when the browser is ready and ends when the connection closes, unless keep_alive is set.
• Busy: if you're at your thread limit (429 threads_exceeded) or the fleet is momentarily busy (503), the connection waits and retries for up to 60 seconds before giving up — client libraries don't retry these themselves.
• Refused: any other error (bad option, plan limits, no key) refuses the connection with the same status and JSON body as the launch request; libraries show it in the connection error (for example 401 Unauthorized).
• The session id comes back in the x-cdpfleet-session-id response header; it's also in the dashboard and GET /v1/me/threads.
• For live links, keep_alive reconnects or anything else that needs the launch response, use the launch request.