Stay free if you only need 1,000 requests per month and access to all quick actions endpoints. Upgrade if you need higher concurrent session and browser-time limits and all quick actions and browser sessions features. Most solo builders can start free.
Why it matters: Browser-time pricing is more expensive than HTTP-only scrapers like ScrapingBee for simple jobs
Available from: Paid (Workers Paid plan)
Why it matters: Several headline features (Live View, Human in the Loop, WebMCP, Stagehand, /crawl) are still in Beta as of April 2026
Available from: Paid (Workers Paid plan)
Why it matters: Debugging serverless browser workflows is harder than running Puppeteer locally
Available from: Paid (Workers Paid plan)
Why it matters: Tightly coupled to the Cloudflare Workers ecosystem for advanced session control
Available from: Paid (Workers Paid plan)
Why it matters: Limits apply to concurrent sessions and total browser time per plan, requiring careful queue design at scale
Available from: Paid (Workers Paid plan)
Why it matters: Get help when stuck. Can save hours of troubleshooting on critical projects.
Available from: Paid (Workers Paid plan)
Browser Rendering runs actual Chrome instances on Cloudflare's edge network, providing true JavaScript execution and DOM manipulation that traditional HTTP scrapers cannot match. This is essential for modern SPAs, complex user interactions, and visual content like screenshots or PDFs. Compared to the 870+ AI tools in our directory, its differentiator is the combination of global edge distribution and a 1,000-request free tier. The trade-off is higher per-request cost than pure HTTP scraping for simple, static pages.
Yes — Cloudflare maintains forks of both Puppeteer and Playwright with compatible APIs, so most existing scripts port over with minimal changes. You can also connect via the raw Chrome DevTools Protocol (CDP) over WebSocket from local machines, CI/CD runners, or external servers, which removes the need to deploy a Worker at all. Some advanced features behave differently in the serverless environment, so test browser launch options, file system access, and long-running session patterns carefully.
The /json Quick Action accepts a URL and a natural language prompt or JSON schema, then returns structured data extracted from the rendered page. Instead of writing CSS selectors that break when sites change, you describe what you want — for example, 'extract product name, price, and rating' — and the underlying model handles the parsing. This is particularly useful for AI agents and data pipelines that need resilient extraction across many different sites without per-site selector maintenance.
Browser Run is available on both Free and Paid plans. The Free tier includes 1,000 requests per month, suitable for development and low-volume use cases. The Paid plan bills based on browser execution time, so simple screenshots or single-page extractions cost a fraction of a cent, while long automation flows that keep a browser open cost more. Reusing sessions via Durable Objects can materially reduce cost for multi-step workflows by eliminating repeated cold starts.
Yes — this is one of its strongest use cases. Browser Run integrates with Playwright MCP and Stagehand so LLM agents can drive browsers using natural language intent rather than CSS selectors, which makes agents more resilient to site changes. The new WebMCP feature (Beta in 2026) and Cloudflare's Agents product extend this with persistent state via Durable Objects and Human-in-the-Loop checkpoints. For autonomous agents that need to log in, navigate flows, and extract data, it is one of the more agent-native browser platforms among the 870+ tools we track.
Start with the free plan — upgrade when you need more.
Get Started Free →Still not sure? Read our full verdict →
Last verified March 2026