outcrawl

Browsers

A real browser for your agent to drive.

Your agent drives it in the cloud through Playwright, Puppeteer or our SDK. Logins carry over, and every session is recorded. 1 credit a minute, about $0.19 to $0.29 an hour.

Get a free API keyRead the browser docs

1,000 free credits every month. No card needed.

The script, as Playwright writes it: import { chromium } from 'playwright'; const browser = await chromium.connectOverCDP( `https://api.outcrawl.ai/connect?token=${process.env.OUTCRAWL_API_KEY}`, ); const context = await browser.newContext(); const page = await context.newPage(); await page.goto('https://practicetestautomation.com/practice-test-login/'); await page.fill('#username', 'student'); await page.fill('#password', 'Password123'); await page.click('#submit'); await page.waitForURL('**/logged-in-successfully/'); console.log(await page.title()); await browser.close();
Session
completed7.1 s0.460823 credits · about $0.0023recorded
  1. 0.7 sconnectOverCDPapi.outcrawl.ai/connect
  2. 1.3 snewPage
  3. 4.4 sgoto/practice-test-login/ 200
  4. 5.3 sfill#username
  5. 5.7 sfill#password
  6. 7.2 sclick#submit
  7. 7.2 swaitForURL/logged-in-successfully/
  8. 7.5 stitleLogged In Successfully | Practice Test Automation
  9. 7.6 sclose

Captured from a real Playwright session on Outcrawl on 7 October 2026, its replay recorded.

How it works

Ask, drive, watch it back. The browser never runs on your machine.

  1. 01

    Your agent connects with one URL.

    Ask Claude Code or Codex for something on a site. It points Playwright or Puppeteer at api.outcrawl.ai/connect with your API key, or calls oc.browser() in our TypeScript SDK, and a browser opens on Outcrawl.

  2. 02

    It drives the browser like any other.

    It opens pages, fills forms, clicks, reads and takes screenshots. The traffic leaves from the country you pick.

  3. 03

    It closes the browser, and you watch it back.

    Closing ends the session and the billing. The session stays in your account with its replay and a log of every request it made.

How browsers work in the docs

Every session

Logins carry over, and every session is on the record. Read it back yourself, or ask your agent to.

  • profile

    Saved logins and cookies. Open the next session with the same profile and it starts signed in.

    oc.browser({ profile })

  • proxy

    Where its traffic leaves from: a country, region or city on our network, such as outcrawl://de/berlin.

    ?proxy=outcrawl://de

  • replay

    A recording of the session to watch, share by link or export to play offline.

    session.replayUrl

  • network

    Every request the pages made, with headers, timings and sizes. Also as a HAR file for your browser's developer tools.

    oc.sessions.network(id)

  • digest

    What went wrong, in short: failed and blocked requests, the slowest ones and console errors.

    oc.sessions.network.digest(id)

  • console

    The console lines the pages printed, with the script and line they came from.

    oc.sessions.console(id)

  • videos

    A video of each tab, one link per recorded stretch.

    oc.sessions.videos(id)

  • record

    On unless you turn it off. Turned off, nothing is kept, and the price is the same.

    record: false

Your agent reads a session with outcrawl_sessions_get, outcrawl_session_network and outcrawl_session_request. The replay is for a person to watch, and it never plays by itself.

What you can build

Your agent scripts what never changes. Real examples you can hand to your agent.

Docs

Playwright for the steps that never change, an agent for the pages that always do.

example
Script the Northgate sign-in with my northgate-ops profile and read the shipment exceptions table. Then send an agent to each flagged shipment for what went wrong and the new delivery date. outcrawl_profile_list: the profile that keeps the portal login. Write(read-exceptions.ts): oc.browser with the profile, the sign-in, the table. Bash(npx tsx read-exceptions.ts): the flagged shipments. outcrawl_agent: one run per shipment · watch it live.
80 blocked sites, run 7 October 2026

Benchmark

It works on the sites where other agents get stopped.

Browser Use built this test from the 80 sites where AI agents got blocked most often in their own traffic. We ran it on Outcrawl and scored it with their own code. The next best result was 81.4%.

Pages reached

92.5%

Mean of 3 runs on 71 scored tasks, 95% interval 91.5% to 94.4%.

  1. Outcrawl92.5%
  2. Browser Use Cloud81.4%
  3. Anchor76.8%
  4. Kernel67.1%
  5. Browserless54.0%
  6. Steel47.2%
  7. Browserbase41.9%
  8. Hyperbrowser39.9%

Mean pass rate, 71 scored tasks, Browser Use’s own scoring code. Ours: 3 runs, 7 October 2026.Method

Pricing

One credit a minute. Billed by the second, plus the data it loads.

Tool

Dollars on

  • Browsers

    A real browser in the cloud that you control.

    1 credit a minute

    about $0.29 an hour

  • Proxy usage

    Data through a proxy, on browser sessions and agent runs.

    2,000 credits a GB

    $9.80 a GB

  • Browser time is billed by the second, from connect to close. An hour is 60 credits, about $0.19 to $0.29.

  • Data the browser loads through our network is 2,000 credits a GB. The sign-in above loaded 172 KB, 0.344156 credits.

  • An hour that loads 100 MB is 60 + 200 = 260 credits, about $0.83 to $1.27.

  • Recording costs nothing extra. Turning it off costs the same.

See full pricing

Options

Three options, and nothing else to set. Everything else matches where the browser is.

Start signed inprofile
Default: none
A saved profile's logins and cookies. One session uses a profile at a time.
Browse from a placeproxy
Default: none
A country, region or city on our network: outcrawl://us/texas/austin.
Keep a recordingrecord
Default: true
On unless your agent sends false.
Every option in the docs

Where browsers stop

FAQ

Questions about browsers