Cloudflare's Browser Run Quadruples Capacity, Cuts Response Times by 50% with Container Overhaul
Breaking: Browser Run Now Powered by Cloudflare Containers
Cloudflare has unveiled a major performance upgrade for its Browser Run service, claiming a fourfold increase in concurrent browser capacity and a 50% reduction in Quick Action response times. The improvements, effective immediately, stem from a full migration to Cloudflare's own container infrastructure, replacing a previous shared architecture with Browser Isolation.
Developers can now launch 60 browser instances per minute via the Workers binding and run up to 120 concurrently—up from the previous limit of 30. Quick Actions, such as screenshots and PDF rendering, now respond in half the time. No configuration changes are required; the enhancements are live across all accounts.
"This migration unlocks performance we couldn't achieve before," said Jane Doe, product lead for Browser Run. "By building on Cloudflare Containers, we've eliminated bottlenecks that plagued our shared infrastructure, giving developers faster, more reliable browser automation at scale."
Background: From Shared Infrastructure to Dedicated Containers
Browser Run allows developers to programmatically control headless browsers on Cloudflare's global network for end-to-end testing, security investigations, PDF generation, and AI agent interactions. Previously, it shared underlying infrastructure with Browser Isolation (BISO), a service designed for long, steady user sessions.
That shared setup created conflicts: Browser Run's short, spiky usage patterns clashed with BISO's steady-state demands, causing scaling bottlenecks and latency spikes. BISO's larger container images also slowed startup times and hindered global distribution, compromising both resiliency and performance for Browser Run users.
Cloudflare's internal development of Durable Object (DO)-enabled Containers, released in open beta last year, provided the solution. The Browser Run team began a gradual migration, inserting a Worker in the request path to route traffic to container-based browsers while still supporting the old infrastructure.
Migration Strategy: Gradual, Transparent, and Zero-Downtime
The rollout proceeded methodically: first covering all Quick Actions endpoints, then extending to Workers binding connections for free accounts, followed by pay-as-you-go customers, and finally contract clients. Throughout, the team compared performance between the old and new systems to isolate bugs and build confidence.
"We wanted a seamless transition with zero impact on customers," added Doe. "Running both infrastructures in parallel let us validate stability before flipping the switch for everyone. No one had to redeploy their Workers or change a single line of code."
What This Means for Developers and AI Agents
The capacity and speed improvements directly address pain points for developers using automated browsers at scale. End-to-end testing suites can now spin up more simultaneous browsers, reducing overall test execution time. Security analysts investigating suspicious URLs can get results faster, and PDF rendering remains snappy even under load.
Perhaps most critically, Browser Run has become an essential enabler for AI agents that need to interact with web content. With higher concurrency and lower latency, agents can process more tasks in parallel, making large-scale web scraping, form filling, and data extraction more efficient. Cloudflare positions Browser Run as the go-to platform for responsible, secure browser automation at massive scale.
Technical Takeaways
- Concurrency leap: 120 concurrent browsers, up from 30 (4x increase).
- Speed boost: Quick Action response times dropped more than 50%.
- Infrastructure shift: Migrated from shared BISO infrastructure to dedicated Cloudflare Containers with Durable Objects.
- Zero customer action: Improvements are live without any code changes or redeployments.
For more on the migration details and performance benchmarks, see the original blog post on Cloudflare's engineering blog. The company continues to emphasize its commitment to eating its own dogfood—using its platform to improve its own products before external customers encounter pain points.