≈100 ms
General Warm Session
Measured warm path for requests without a supplied detailed fingerprint. Best for general scraping, stateless agents, and fast testing.
Benchmarks
Benchmark browser startup the way your agents use it.
Browser startup depends on the path your request takes: a roughly 100 ms general warm session, a roughly 150 ms warm session with a supplied detailed fingerprint, or a cold fallback when fresh capacity is needed. This page explains how to evaluate those paths without treating one lab result as a universal promise.
Startup paths
"Warm" is not just one path.
We keep browsers pre-warmed so general requests can start in roughly 100 ms. Supplying a detailed fingerprint measures around 150 ms, while custom state, proxy configuration, or a cold fallback can still change startup behavior.
≈150 ms
Supplied-Fingerprint Warm Session
Measured warm path when you supply a detailed fingerprint. Storage-heavy or proxy-specific payloads can still add workload-specific setup.
Cold fallback
Cold Start Fallback
When demand temporarily outpaces warm capacity, BrowserCity can fall back to a fresh runtime. This path is slower and depends on workload, region, and browser setup.
Methodology
Beyond "Time to First Byte"
A useful benchmark measures the full journey and names its environment, date, BrowserCity version, sample size, and distribution. Use that context before comparing startup paths or setting customer-facing expectations.
From request to readiness
Useful startup measurement should cover the whole journey: request admission, browser allocation, identity or storage preparation, and the point where the remote browser is ready to navigate via MCP or Playwright.
Accounting for identity setup
Warm starts are not all equivalent. General warm requests are roughly 100 ms, while supplied detailed fingerprints are roughly 150 ms. Storage injection, proxy setup, and other custom requirements can add work before handoff.
Cold fallbacks
Warm pools absorb normal bursts, but a drained pool or uncommon configuration can require a fresh browser runtime. Treat cold fallback as workload-dependent and measure it against your own traffic pattern.
Measure your workload
Test the startup path your agents will use.
Connect Playwright, your MCP client, or trigger a REST action with and without supplied detailed fingerprints, storage state, and your real traffic pattern to see how BrowserCity behaves for your workload.
[ 06 / 06 ] — Get Started
Give your AI agents the web.
Start for free. No credit card required. Private sessions by default.