Compare / other services
cdpfleet vs Oxylabs Agent Browser
A fleet of real browsers you drive with Playwright, paid per browser-second, against Oxylabs' Agent Browser (formerly Unblocking Browser): a managed Chromium with residential proxies, paid per GB — and its Web API search and scrape endpoints against cdpfleet Functions. Prices, limits and features side by side.
Oxylabs' Agent Browser (formerly Unblocking Browser, also sold as a headless browser service) is a managed browser in Oxylabs' cloud — Chromium or Oxylabs' own engine: you attach Playwright or Puppeteer over CDP with your credentials in the URL, and set the exit country, device, fingerprint, challenge handling, recording or a saved profile with query parameters. Oxylabs' residential proxies are built in. You pay by the GB, out of one credit balance shared with Oxylabs' other products. This page compares the Agent Browser with cdpfleet sessions, and Oxylabs' Web API (search and scrape) and Web Unblocker with cdpfleet Functions, jobs and monitors; its proxy products are left out. cdpfleet is a fleet of 12 real desktop browsers launched with one HTTPS call and driven with the Playwright protocol from five languages: you choose the browser and the proxy, and pay per browser-second.
| cdpfleet | Oxylabs Agent Browser | |
|---|---|---|
| Browsers | 12 — Google Chrome, Microsoft Edge, Brave, Opera, Naver Whale, Yandex Browser, Chromium, Patchright, Rebrowser, Camoufox, Firefox, WebKit, plus hardware Android phones and iOS WebView in early access. Chrome with stable, beta and dev channels and pinned previous majors. | Two engines: standard Chromium (default) or Oxylabs' proprietary engine (p_src=oxybrowser). Windows (default) or macOS fingerprint profiles (p_fp_tag), a desktop (default) or mobile device (p_device); no version choice documented. |
| How you connect | Playwright protocol: connect(wsUrl) with the matching Playwright client from Node.js, Python, Java, C# or Go — all contexts you like. Or "cdp": true on Chromium-based browsers for Puppeteer and connectOverCDP (CDP & Puppeteer). | CDP at wss://USER:[email protected] — Playwright connectOverCDP (examples in Python and Node.js), Puppeteer, Stagehand, browser-use or any CDP tool. Options go in the URL as query parameters (?p_cc=US&page_handler=true); a proxy can't be set per context. Default limits: 100 concurrent sessions and 10 new sessions per second; Oxylabs recommends retrying the connect with backoff. |
| Session length | Up to 24 h (shared threads: also capped by your remaining daily minutes). Idle timeout 120 s by default, up to 10 min on shared threads and 24 h on dedicated. Any number of sites per session. | Named sessions (session_name) live up to 24 h, 5 of them at a time by default; with keep_alive=true one stays up to 10 minutes after you disconnect. An idle timeout or a maximum length for ordinary sessions isn't documented. |
| Metering | Per second, as thread-seconds: a headless browser is 1 thread, headful 2. Billing stops when the WebSocket closes. Traffic through your own proxies is free. Prices per thread are on the pricing page. | Per GB of request and response traffic, "regardless of the HTTP status code returned", at your plan's rate: $8/GB on the smallest plans down to $4/GB on Enterprise. The pricing page lists no charge for browser time, and all features — geo-targeting, CAPTCHA handling — are included. |
| Plans | Free Community plan: 2 threads, 60 browser-minutes per thread a day, 15-minute sessions, proxies included. Paid from $7.50/month (5 shared threads, 60–480 browser-minutes a day per thread); dedicated threads for 24/7 use at $8.50/thread/month. Pricing. | One credit-based subscription for all Oxylabs products, spent at each product's rate; unused credit expires at the end of each cycle, and the service pauses when the balance runs out (top-ups of any amount). Free $5 welcome credit (up to 0.6 GB of Agent Browser traffic at $8/GB). Starter $15/month ($8/GB, up to 2 GB), Basic $100 ($7.5/GB, up to 13 GB), Flexible $500 ($5.5/GB, up to 90 GB), Enterprise from $2,500 ($4/GB, 625+ GB). Self-service since 23 October 2026. |
| Concurrency | The threads you buy: 5 to 10,000 (more on request), 2 on the free Community plan; each session takes 1–4 threads by type. | 100 concurrent sessions and 10 new sessions per second by default on every plan; more through support. Up to 1,000 open pages per session. |
| Proxies | Bring your own HTTP, HTTPS or SOCKS5 proxies — single, pool, or per-host rules (proxy_rules) — swap a running session's proxy live, see traffic per destination host and per rule; traffic through them is free. Or the cdpfleet residential proxy: 4 tiers from $0.50 to $5/GB from one prepaid balance ($5 free credit), mixable per host — e.g. Tier 4 for CDN assets, Tier 1 for pages. A launch without a proxy uses the free Community proxy — best effort, no guarantees. | Oxylabs residential proxies built in: a new exit IP per connection, or a pinned one for 1 to 1,440 minutes (proxy_resi_ses_id, proxy_resi_ses_time). Country (p_cc), US state (p_state) or city (p_city, best effort). Only residential is documented; using your own proxy isn't. |
| Stealth | Real browsers by default (genuine Chrome, Edge, Brave…), headful by default, and three stealth builds — Patchright, Rebrowser, Camoufox — on every plan. You decide what the site sees. | Managed fingerprints, an engine and OS profile to choose (for protected targets Oxylabs suggests its own engine with "macOS optimized" fingerprints), a desktop or mobile device. Page Handler (off by default, enabled per account) identifies each page, handles the challenges it supports and reports the page state — about 8 s on a page without a challenge, typically 30–45 s with one; the older solve_captcha=true is being phased out. |
| What you may do | Log in, type passwords, set headers and cookies, visit as many domains as you like in one session — within the terms. | Restricted by default: streaming, banking and payments, government, gaming, ticketing, webmail, ad and IP-lookup sites; more categories open after a KYC check with your account manager or support. |
| Watching a session | Live view with human takeover — Watch / Take control in the dashboard, or a link from the API — and Launch interactive: open any browser on a start page straight into live view from the dashboard's code generator (start_url over the API). Session recordings as MP4 (480p free, 720p or 1080p on a plan), and JPEG frames to your code with page.screencast. | Live view over VNC from the Agent Browser dashboard, or a viewer link per session ID; recordings with record=true (a named video, one continuous file across reconnects of a named session), up to 10 kept by default. |
| Persistence | Persistent profiles keep cookies, local storage and IndexedDB between sessions (paid plans); Playwright storageState; keep_alive to reconnect to the same browser on dedicated threads. | Persistent profiles: o_profile saves and restores cookies and localStorage by name (o_profile_save forces a save), kept 14 days from last use; 5 profiles by default. Oxylabs advises pinning the same exit IP when restoring a profile. |
| Scraping in one call | Functions: screenshot, PDF, Markdown, scrape (HTML, Markdown, text, links, screenshot, XHR calls, cookies; wait_for, actions, extract by CSS selector), map, geo compare and a live tab — in a pooled browser, or in a browser and device of your choice on paid plans. Billed like a session (one thread while the call runs, one per country for geo compare); the Community proxy and your own proxies cost nothing extra. Free tools at /functions. | Web API /v1/scrape (new in October 2026): one URL in; Markdown, HTML, a PNG screenshot or JSON — from a dedicated parser, or AI parsing to a prompt or a JSON Schema — with a country, desktop or mobile, and run_js for a headless browser (Oxylabs suggests a 150-second client timeout). No action steps documented on it. $1.50 per 1,000 results with JavaScript and $1.00 without on Starter, $1.22 and $0.75 on Enterprise; AI parsing $3 per 1,000 pages; from the shared balance. Web Unblocker, a proxy-style endpoint billed per GB of traffic (not for its own 5xx failures), renders HTML or a PNG on request and runs browser instructions (beta) — type, click, wait, or return a matching XHR response. |
| Crawling and bulk | Jobs on paid plans: batch (any number of URLs), crawl and map, with no size or time limit — runs continue past the 24-hour session limit by themselves; results as JSON lines or a zip archive, kept 14 days; stop and resume; signed webhooks. Plus watch (live XHR/WebSocket/DOM events), warmup and monitors (scheduled change detection with diffs). | No crawl or map endpoint in the Web API: asynchronous jobs report to a callback_url or upload to S3, GCS or S3-compatible storage, and a batch sends up to 5,000 URLs in one request (generic /scrape and some dedicated scrapers). No schedules or change detection in the Web API docs. The older Web Scraper API — now "Classic", with credentials from Oxylabs' team — has a free Scheduler (a cron expression, a set of jobs and an end time); its Web Crawler page is gone from the docs. |
| Screenshots and PDFs | Screenshot (PNG/JPEG, viewport or full page) and PDF Functions — a text PDF from a Chromium-based browser. | Web API /scrape returns a base64 PNG of the fully rendered page (run_js: true required); Web Unblocker a PNG with X-Oxylabs-Render: png. In an Agent Browser session, your own code's screenshots; the self-hosted Agent Browser MCP server has screenshot and PDF-export tools. No PDF endpoint. |
| Search and site APIs | None — no search-engine or per-site APIs; a scrape or a session reads whatever page you point it at. | Web API /v1/search: up to 20 ranked organic results with snippets, "People also ask" and related searches, by country, $1 per 1,000 searches; a search that fails on every engine isn't charged. Dedicated scrapers (about 160 endpoints) return parsed JSON for Google (search, ads, AI Overviews, AI Mode, Shopping, Maps, Lens, Trends, Hotels), Bing, Amazon, YouTube, ChatGPT, Gemini, Perplexity, Walmart, eBay and many other retailers. |
| Agent tooling | A hosted MCP server at cdpfleet.com/mcp — add it in Claude, ChatGPT or Gemini and sign in with OAuth — or npx cdpfleet-mcp; SDKs for Node.js, Python and Go. | For the browser, a self-hosted MCP server (hb-mcp, installed from GitHub) with session, Page Handler, recording and profile tools plus navigate, snapshot, click, type, screenshot and PDF export; its sessions last up to an hour. For the Web API, a hosted MCP server (mcp.oxylabs.io/web-api-mcp, Bearer API key — clients limited to OAuth can't connect) or a self-hosted one, with search, scrape and dedicated-scraper tools. Agent Skills for Claude Code, Cursor and other assistants; Stagehand and browser-use connect over CDP. No SDK of its own for the browser. |
| Examples | 89 cases run on production in five languages, all on GitHub with their output. | Connection, retry, CAPTCHA, Page Handler and profile examples in Python and Node.js (Playwright, Puppeteer) in the docs; Stagehand and browser-use examples. |
Oxylabs Agent Browser facts from its public pricing page and documentation, checked on 12 November 2026; list prices in USD. Tell us if something is out of date.
Which one, when
Pick Oxylabs Agent Browser if…
- You want residential proxies, fingerprints and challenge handling bundled with the browser, and only write the scraping logic.
- Your pages are light and your sessions long or idle: you pay per GB, not for browser time.
- You want web search and parsed results from Google, Amazon or AI assistants in one API, or already use Oxylabs proxies and want one balance for all of them.
Pick cdpfleet if…
- The site must see a specific real browser or version — Chrome channels, Edge, Brave, Opera, Firefox, WebKit, Camoufox — or a phone.
- You bring your own proxies, free of traffic charges, or want residential traffic from $0.50/GB — useful on heavy pages, where per-GB pricing adds up.
- You drive the browser from Java, C# or Go with the full Playwright protocol, or need more than 100 sessions at once without asking.
- You want scrapes with actions and PDFs in one call, crawls that follow links, and monitors that keep diffs — next to the browser, from one API.
- You want to start free — the Community plan includes proxies — and try a browser in live view before writing any code.
Try it
The quickstart is one HTTPS call and five lines of Playwright; the code builder writes the launch for any browser and option set in your language.
Sources
- Oxylabs Agent Browser pricing
- Agent Browser product page (Chromium, MCP, residential proxies)
- Agent Browser documentation (connection, parameters, limits, profiles)
- Quick start: Agent Browser
- Agent Browser engine and fingerprints
- Agent Browser Page Handler
- Agent Browser geolocation and proxy selection
- Agent Browser CAPTCHA handling
- Agent Browser session inspection and recording
- Agent Browser persistent sessions and profiles
- Agent Browser response codes (limits)
- Agent Browser troubleshooting (rate limits, KYC, defaults)
- Agent Browser restricted targets
- Oxylabs credit-based billing
- Oxylabs Web API
- Web API /scrape
- Web API /search
- Web API delivery modes (async, webhooks, batch)
- Web API dedicated scrapers
- Web API AI parsing
- Web API for agents (errors and billing of failed searches)
- Web API product page and pricing
- Web Unblocker JavaScript rendering
- Web Unblocker browser instructions (beta)
- Web Unblocker billing
- Web Scraper API (Classic)
- Web Scraper API Scheduler
- Oxylabs MCP servers (Web API, Agent Browser)
- Oxylabs Agent Skills
- Agent Browser CDP integration (Stagehand, browser-use)
- Oxylabs changelog