At a glance
Published by Notte, which competes with the products on this page. Every figure links to the provider's own documentation, checked on 2026-11-02.
| Item | Value |
|---|---|
| Availability | Self-serve[1] |
| Billing unit | Browser hours for all methods: usage is totaled in seconds per day and the monthly total is rounded to the nearest whole hour. Browser Sessions (Puppeteer, Playwright, CDP) are also billed for concurrent browsers, calculated as the monthly average of the daily peak.[2] |
| Free tier | Workers Free: 10 minutes of browser time per day and up to 3 concurrent browsers (Browser Sessions).[2] |
| Entry paid plan | Workers Paid ($5/mo)[2] |
| Browser-hour rate (self-serve plans) | $0.09[2] |
| Max concurrency | Workers Free 3 concurrent browsers; Workers Paid 200 per account by default, with higher limits available on request.[3] |
| Max session duration | No fixed maximum lifetime while the session stays active. The inactivity timeout can be raised to 20 minutes, so a session must receive a command inside that window to stay open. Sessions also close when Browser Run rolls out a new release.[3] |
| Idle timeout | 60 seconds of inactivity by default; the keep_alive option extends it to up to 20 minutes.[3] |
| Regions | Cloudflare's edge network, nearest location chosen automatically; no region parameter found[1] |
| SDK languages | typescript, javascript, python, go[4] |
| Open source | Partial: Apache-2.0 (Puppeteer and Playwright client forks), github.com/cloudflare/puppeteer[5] |
| SOC 2 | Yes: Cloudflare's SOC 2 Type II report covers security, confidentiality and availability controls and lists Browser Rendering (the former name of Browser Run) under Developer Platform as in scope. The report is shared under a nondisclosure agreement.[6] |
| HIPAA | Enterprise: Browser Rendering is listed as an in-scope Developer Platform service for Cloudflare's HIPAA AoC. Cloudflare signs a BAA only with Enterprise-level customers.[7] |
| GDPR and DPA | Yes: Cloudflare publishes a standard Data Processing Addendum that forms part of its Enterprise Subscription Agreement, Self-Serve Subscription Agreement or other written or electronic agreement, and defines the EU SCCs and the Data Privacy Framework for transfers. Company-wide terms, not specific to Browser Run.[8] |
Sources: developers.cloudflare.com [1] ↗developers.cloudflare.com [2] ↗developers.cloudflare.com [3] ↗developers.cloudflare.com [4] ↗developers.cloudflare.com [5] ↗cloudflare.com [6] ↗cloudflare.com [7] ↗cloudflare.com [8] ↗
Pricing
| Plan | Monthly | Included | Browser-hour rate | Concurrency | Max session | Proxy per GB | Retention |
|---|---|---|---|---|---|---|---|
| Workers Free[1] | $0 | Not published | Not published | 3 | Not published | Not published | Not published |
| Workers Paid[2] | $5/mo | 10 hours | $0.09 | 200 | Not published | Not published | Not published |
- Workers Free, as published: 10 minutes per day of browser time; 3 concurrent browsers; no overage (429 until next UTC day)[1]
- Workers Paid, as published: 10 hours per month, then $0.09 per additional hour; 10 concurrent browsers (averaged monthly), then $2.00 per additional browser[2]
Sources: developers.cloudflare.com [1] ↗developers.cloudflare.com [2] ↗
Proxy, CAPTCHA and other meters
The notes below were recorded with the sourced values. They are not quotations from the provider.
| Meter | Pricing |
|---|---|
| Proxy | No proxy product. Browser Run traffic comes from Cloudflare IP ranges and per-request IP rotation cannot be configured.[1] |
| CAPTCHA solving | No CAPTCHA solving product. Browser Run does not bypass CAPTCHAs or other bot protection.[2] |
| Concurrent browsers (Browser Sessions only) | Workers Paid: 10 included (averaged monthly), then $2.00 per additional browser. Concurrency is the monthly average of the daily peak number of concurrent browsers.[3] |
| Workers AI usage for /json | The /json endpoint (and the JSON format of /crawl) uses Workers AI by default and incurs Workers AI usage, billed separately from browser hours. A custom model with your own API key can be used instead.[4] |
| /crawl with render: false | render: true crawls bill under Browser Run pricing. render: false crawls run on Workers, were not billed during the beta and will later bill under Workers pricing.[2] |
| Kitesurf browser backend | Free while in beta, behind per-account limits.[5] |
- Browser Run limits and pricing follow the Cloudflare Workers plan (Free or Paid). The $5/month Workers Paid minimum covers Workers and related products, so a Browser Run bill is that minimum plus overage.
- Billing time: usage is totaled in seconds per day, summed over the cycle and rounded to the nearest whole hour (1,800 seconds or more rounds up, 1,799 or less rounds down). A failed Quick Actions request that hits waitForTimeout is not charged.
- Quick Actions are billed for browser hours only. Every Quick Actions response returns the X-Browser-Ms-Used header with browser milliseconds used.
- Example from the pricing page: 50 browser hours plus 15 average concurrent browsers on Workers Paid = (50 - 10) x $0.09 + (15 - 10) x $2.00 = $13.60 on top of the plan minimum.
- Browser close reasons and daily usage are visible in the dashboard Browser Run page; sessions not closed with browser.close() keep billing until the inactivity timeout.
Sources: developers.cloudflare.com [1] ↗developers.cloudflare.com [2] ↗developers.cloudflare.com [3] ↗developers.cloudflare.com [4] ↗developers.cloudflare.com [5] ↗
Limits
| Limit | Value |
|---|---|
| Max session duration | No fixed maximum lifetime while the session stays active. The inactivity timeout can be raised to 20 minutes, so a session must receive a command inside that window to stay open. Sessions also close when Browser Run rolls out a new release.[1] |
| Default session timeout | 60 seconds of inactivity[1] |
| Idle timeout | 60 seconds of inactivity by default; the keep_alive option extends it to up to 20 minutes.[1] |
| Max concurrency | Workers Free 3 concurrent browsers; Workers Paid 200 per account by default, with higher limits available on request.[1] |
| Session creation rate limit | Browser Sessions: 1 new browser every 20 seconds on Free, 3 per second on Paid. Quick Actions: 1 request every 10 seconds on Free, 30 per second on Paid. Excess requests get HTTP 429 with a Retry-After header.[1] |
| Regions | Cloudflare's edge network, nearest location chosen automatically; no region parameter found[2] |
Sources: developers.cloudflare.com [1] ↗developers.cloudflare.com [2] ↗
Features
| Feature | Group | Status | Detail |
|---|---|---|---|
| Connect over CDP | Connect | Yes[1] | wss://api.cloudflare.com/client/v4/accounts/{account_id}/browser-run/devtools/browser with a Bearer API token; any CDP client can connect. HTTP session endpoints create sessions and list or open tabs. keep_alive, lab, recording and browser query parameters are supported. |
| Playwright | Connect | Yes[2] | Playwright through Cloudflare's fork @cloudflare/playwright inside Workers (browser binding), or stock playwright-core via connectOverCDP from any Node.js environment. Playwright MCP server fork is beta. |
| Puppeteer | Connect | Yes[3] | Puppeteer through Cloudflare's fork @cloudflare/puppeteer inside Workers (browser binding), or stock puppeteer via CDP from any Node.js environment. |
| Selenium | Connect | Unknown | No Selenium or WebDriver support is mentioned in the Browser Run docs (searched llms-full.txt). |
| Persistent profiles | Login and state | Unknown | No cookie, profile or user-data-dir persistence feature found. Sessions can be reused by reconnecting for up to a 20 minute inactivity window, or kept alive with Durable Objects, and isolated with browser contexts. |
| Credential vault | Login and state | Unknown | No credential vault or secret-filling feature found. Human in the Loop is the documented route for credentials the script should not handle. |
| Managed login | Login and state | Unknown | No managed login or auth-agent feature found. Quick Actions accept HTTP credentials (authenticate), cookies and extra headers. |
| TOTP and 2FA handling | Login and state | Unknown | No TOTP or MFA code handling found; docs route MFA logins to Human in the Loop. |
| CAPTCHA solving | Anti-bot and network | No[4] | Browser Run does not bypass or solve CAPTCHAs, Turnstile or other bot protection. Human in the Loop is the documented route for CAPTCHAs. |
| Stealth and fingerprinting | Anti-bot and network | Unknown | No stealth or fingerprint management feature is documented. Cloudflare states that Browser Run requests are always identified as bot traffic and include automatic headers so origin servers can identify them. |
| Managed proxies | Anti-bot and network | No[5] | No provider proxies. Outbound traffic comes from Cloudflare IP ranges and per-request IP rotation cannot be configured. |
| Proxy geo-targeting | Anti-bot and network | Unknown | No exit-country or city selection is documented. The FAQ says all rendering traffic comes from Cloudflare IP ranges and per-request IP rotation cannot be configured. |
| Bring your own proxy | Anti-bot and network | Unknown | No bring-your-own-proxy option found. Outbound Workers can route selected hostnames through another Worker (for private services, authentication or response processing), which is not a general proxy setting. |
| Live view | Observe and control | Beta[6] | Watch and interact with an active Browser Session from the dashboard Live Sessions tab, a generated devtoolsFrontendUrl on live.browser.run (tab, full or devtools modes), or native Chrome DevTools. URLs carry a signed jwt, are valid 5 minutes to start a connection by default (up to 1 hour), and can be made read-only. Available for Browser Sessions (Puppeteer, Playwright, CDP). |
| Embeddable live view | Observe and control | Unknown | Live View URLs can be opened in any browser and shared as read-only links, but iframe embedding in your own app is not documented. |
| Human takeover | Observe and control | Beta[7] | Human in the Loop: a human steps into a live session through Live View. A structured handoff uses Cloudflare.handoff and the Cloudflare.handoffComplete CDP events (instructions for the operator, Done or Failed buttons, optional timeout up to 30 minutes, added 2026-07-28), or the script can poll for completion. |
| Session recording | Observe and control | Beta[8] | Opt-in rrweb recording (recording: true on launch or recording=true on the CDP URL), stored as structured JSON events rather than video, replayed in the dashboard Runs tab. Retained 30 days, 1 second minimum and 2 hours maximum duration. Browser Sessions only, not Quick Actions. |
| Network logs | Observe and control | Beta[8] | Session Recordings capture network activity per tab target. The API returns a JSON array of the recorded requests (available request and response headers, payloads, status codes, timing, transfer sizes), or an HTTP Archive (HAR) 1.2 document with format=har. |
| File downloads | Files and extensions | Unknown | No file download handling is documented for Browser Run. |
| File uploads | Files and extensions | Unknown | No file upload handling is documented for Browser Run. |
| Browser extensions | Files and extensions | Unknown | No browser extension loading is documented for Chromium sessions. The Kitesurf backend is described as trading away extensions. |
| Serverless functions | Run code and agents | Unknown | Browser Run has no function-hosting product of its own. It is driven from Cloudflare Workers, a separate Cloudflare product, through a browser binding. On Workers Paid, CPU time per HTTP request is up to 5 minutes (default 30 seconds). |
| Scheduling | Run code and agents | Unknown | Browser Run documents no scheduling feature of its own. Workers Cron Triggers, a separate Workers feature, can invoke a Worker that launches Browser Run on a schedule. |
| Webhooks | Run code and agents | Beta[9] | Crawl jobs can publish started, updated and finished events to a Cloudflare Queue through a Queues event subscription with source browserRun. No HTTP webhook callback was found. The crawl endpoint is beta. |
| MCP server | Run code and agents | Yes[10] | Remote Cloudflare MCP server for Browser Run at https://browser.mcp.cloudflare.com/mcp (fetch pages, convert to markdown, screenshots). Browser Run also supports chrome-devtools-mcp through CDP and a deployable Playwright MCP fork (beta). |
| Hosted agent API | Run code and agents | Unknown | No endpoint that takes a natural-language task and runs the agent loop was found. The /json endpoint extracts structured data from a prompt with an LLM. Stagehand (beta) and Playwright MCP (beta) are deployed by you in a Worker. |
| Scrape and fetch API | Run code and agents | Yes[11] | Quick Actions: /content, /screenshot, /pdf, /markdown, /snapshot, /accessibilityTree, /scrape, /json (AI extraction), /links over REST or env.BROWSER.quickAction() in Workers. /crawl is beta and REST only. |
| Self-hosting | Deployment | Unknown | Browser Run is a managed Cloudflare service. The Puppeteer and Playwright forks are open source (Apache-2.0); local wrangler dev runs a local Chrome. No self-hosted Browser Run server found. |
| Region selection | Deployment | Unknown | Docs say sessions open close to your users on Cloudflare's global network; no region or location parameter was found. |
Sources: developers.cloudflare.com [1] ↗developers.cloudflare.com [2] ↗developers.cloudflare.com [3] ↗developers.cloudflare.com [4] ↗developers.cloudflare.com [5] ↗developers.cloudflare.com [6] ↗developers.cloudflare.com [7] ↗developers.cloudflare.com [8] ↗developers.cloudflare.com [9] ↗developers.cloudflare.com [10] ↗developers.cloudflare.com [11] ↗
Get started
SDK languages: typescript, javascript, python, go. The snippet below is the official javascript quickstart from developers.cloudflare.com, trimmed.
- javascript: @cloudflare/puppeteer[2]
- javascript: @cloudflare/playwright[3]
- typescript: cloudflare (official API SDK, client.browserRendering)[4]
- python: cloudflare (official API SDK, browser_rendering resource)[5]
const { chromium } = require("playwright-core");
const ACCOUNT_ID = process.env.CF_ACCOUNT_ID || "<ACCOUNT_ID>";
const API_TOKEN = process.env.CF_API_TOKEN || "<API_TOKEN>";
const browserWSEndpoint = `wss://api.cloudflare.com/client/v4/accounts/${ACCOUNT_ID}/browser-run/devtools/browser?keep_alive=600000`;
async function main() {
const browser = await chromium.connectOverCDP(browserWSEndpoint, {
headers: {
Authorization: `Bearer ${API_TOKEN}`,
},
});
const context = browser.contexts()[0];
const page = context.pages()[0] || (await context.newPage());
await page.goto("https://developers.cloudflare.com");
const title = await page.title();
console.log(`Page title: ${title}`);
await page.screenshot({ path: "screenshot.png" });
await browser.close();
}
main().catch(console.error);
Sources: developers.cloudflare.com [1] ↗npmjs.com [2] ↗npmjs.com [3] ↗npmjs.com [4] ↗pypi.org [5] ↗
Recent launches
- 2026-10-30: Chrome 152 default browser, Quick Actions User-Agent update, /links absolute URLs, Markdown keeps code languages[1]
- 2026-09-29: Multiple concurrent client connections to one Browser Run session[1]
- 2026-09-25: Crawl job lifecycle events published to Cloudflare Queues[1]
- 2026-09-18: Inspect panel in Session Recordings: console logs, network requests with HAR export, DOM[1]
- 2026-09-14: Guardrails: restrict a session to permitted hostnames (beta)[1]
- 2026-08-20: Workers Paid default limits raised to 200 concurrent browsers, 3 new browsers per second, 30 Quick Actions requests per second[1]
- 2026-08-06: Kitesurf, an agent-first stateless browser backend (beta, free)[1]
- 2026-07-28: Structured handoff (Cloudflare.handoff) for Human in the Loop[1]
Sources: developers.cloudflare.com ↗