MIT-licensed browser-agent library for Python and TypeScript, plus a pay-as-you-go cloud at $0.02 per browser-hour with CAPTCHA solving included.
Builders who want the agent loop itself, not just a browser: the MIT-licensed library drives Chromium from a task description and runs locally, in the cloud, or as a CLI skill that gives an existing agent browser access. It is the strongest pick when you need published evidence before committing, because Browser Use is the only vendor here that posts a strict task accuracy figure and an anti-bot bypass figure with task counts and dates. Pay-as-you-go with no subscription suits spiky workloads.
Anyone who needs a predictable monthly invoice or guaranteed concurrency should look elsewhere. Concurrency is not sold, it is earned by lifetime spend: 10 sessions at $0, 50 at $200, 250 at $1,000, 500 at $5,000 and 1,000 at $25,000, so a new account cannot run a large fleet on day one at any price. Model tokens through the gateway carry a 20% markup on top of provider rates, bring-your-own-key still pays the 20% fee, and residential proxy is on by default at $5/GB. The README also states plainly that no browser configuration guarantees every CAPTCHA can be avoided or solved.
Pros
Cons
Two, both self-run. Internal Bench Hard reports 82% strict accuracy on 106 hard live-website tasks at $0.17 per solved task, updated 2026-08-01, with credit given only when the final answer or end state is exactly right. The BU Stealth Benchmark reports 81% bypass across 71 anti-bot-protected sites at 100 attempts per provider, updated 2026-03-21, alongside 84.8% on BrowserBench.
The MIT-licensed library is complete and runs against a local Chromium, so you can self-host the agent loop entirely. You still pay a model provider for inference. The cloud adds stealth Chromium, in-browser CAPTCHA solving, residential proxies in 195+ countries and managed concurrency at $0.02 per browser-hour. Teams hitting anti-bot walls locally are the ones who need the hosted browsers.
By lifetime spend, not by plan. The pricing page publishes 10 concurrent sessions at $0 spent, 50 at $200, 250 at $1,000, 500 at $5,000 and 1,000 at $25,000, with higher limits by contacting sales. Limits never go down once reached. A new account therefore starts at ten parallel browsers regardless of willingness to pay, which constrains same-day large batch jobs.
Three meters stack. Browser time is $0.02 per hour billed to the minute. Model tokens cost provider rates plus a 20% fee, applied even when bringing your own key. Residential proxy is on by default at $5 per GB, dropping to $0.20 per GB for direct connections or your own proxy. For most workloads proxy bandwidth and tokens, not browser hours, dominate the invoice.
Facts sourced from the vendor’s own site, last verified 2026-10-10.
The Weekly Source
One email a week. New listings, repriced tools, and the ones we would skip - including this one if its pricing changes.
