BU pricing guide details
Pricing Overview
BU (Browser Use) prices on consumption with no subscription layer underneath it. You start free, then buy prepaid credits from $5, and those credits do not expire. There is no seat count, no monthly platform fee, and no annual commitment gating access to the product — an unusual structure in a category where most agent platforms sell a monthly workspace subscription with a credit allowance attached.
The bill has three independent meters:
- Browser time — $0.02 per hour of browser runtime.
- Internet traffic — billed separately from browser time, at $5 per GB on high-quality residential proxies or $0.20 per GB on a direct connection or your own proxy. Proxies are on by default.
- Agent model usage — Browser Use Agents are billed at model cost + 20%.
The practical consequence is that browser time is almost never the expensive part. At $0.02 per hour, a browser running continuously for a 730-hour month costs $14.60. The variables that actually move the invoice are traffic — where the residential-proxy rate is 25× the direct rate — and model spend, which BU passes through with a 20% margin.
Enterprise pricing is available through sales, and BU lists SOC 2 Type II, HIPAA and zero data retention. BU does not publish an Enterprise price.
Plan Comparison
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Plan | Price | What the data states |
|---|---|---|
| Free start | $0 | Start free, pay as you go |
| Credits (pay as you go) | From $5 | Prepaid credits, no subscription, no expiry |
| Browser runtime | $0.02 / hour | Plus internet traffic |
| Traffic — residential proxy | $5 / GB | High-quality residential proxies; proxies on by default |
| Traffic — direct or BYO proxy | $0.20 / GB | Direct connection or bring-your-own proxy |
| Browser Use Agents | Model cost + 20% | Plus browser time and traffic |
| Enterprise | Talk to sales | More concurrency on request; SOC 2 Type II, HIPAA, zero data retention |
BU also publishes separate references for model prices, concurrency limits, billing details, and a custom model API. Concurrency starts at 10 sessions shared between agents and browsers, rises with total paid per project, and never goes down.
Hidden Costs and Considerations
Proxies are on by default. This is the single most important line item to check before you scale. The default path bills traffic at $5 per GB; switching to a direct connection or your own proxy bills at $0.20 per GB. That is a 25× difference on a meter you may not be watching, because it is billed separately from the browser-hour rate that dominates the marketing copy. If your targets do not require residential IPs, the default setting is costing you 25 times the direct rate. If they do — anti-bot-protected commerce sites, geo-restricted content, sites that block datacenter ranges — the residential rate is the cost of the job working at all, not an optional upgrade.
Traffic volume is hard to predict from task count. Browser automation traffic depends on page weight, not on how many tasks you run. A workflow that loads image-heavy pages, video, or ad-laden marketplaces will consume dramatically more GB per task than one hitting a lean internal admin panel. We would strongly advise measuring GB-per-task on your real targets during the free tier before modelling any monthly figure.
The 20% model markup compounds with retries. Agents that retry, re-plan, or loop on a failing selector consume tokens each attempt, and every token carries the markup. Verbose reasoning models and large screenshot payloads push this meter hardest. If you already hold model API contracts, the custom model API is worth investigating as a way to route model spend outside the markup — though the data we hold does not state whether that changes the 20% treatment.
Prepaid credits are a working-capital question, not a cost. Credits do not expire, so unused balance is not lost, but it is cash out the door ahead of usage. For teams with procurement processes that dislike prepaid balances, that is a process friction rather than a price problem.
Costs BU does not charge that you will still pay. Engineering time to build and maintain the automations, observability and alerting for failed runs, and storage for any artifacts you extract all sit outside this pricing. For usage-based agent platforms generally, the operational cost of keeping brittle automations working typically exceeds the platform bill.
Cost Estimates by Team Size
These are our scenarios, not BU quotes. Each row shows the unit rate and the quantity we assumed so you can substitute your own numbers.
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Scenario | Browser time | Traffic | Model | Monthly total |
|---|---|---|---|---|
| Single developer evaluating | 50 hrs × $0.02/hr = $1.00 | 5 GB × $5/GB residential = $25.00 | $50 model cost + 20% = $60.00 | $86.00 |
| Same developer, BYO proxy | 50 hrs × $0.02/hr = $1.00 | 5 GB × $0.20/GB direct = $1.00 | $50 model cost + 20% = $60.00 | $62.00 |
| Small team, production workflows | 500 hrs × $0.02/hr = $10.00 | 50 GB × $5/GB residential = $250.00 | $500 model cost + 20% = $600.00 | $860.00 |
| Same team, BYO proxy | 500 hrs × $0.02/hr = $10.00 | 50 GB × $0.20/GB direct = $10.00 | $500 model cost + 20% = $600.00 | $620.00 |
| High-volume extraction | 5,000 hrs × $0.02/hr = $100.00 | 500 GB × $5/GB residential = $2,500.00 | $2,000 model cost + 20% = $2,400.00 | $5,000.00 |
| Same volume, BYO proxy | 5,000 hrs × $0.02/hr = $100.00 | 500 GB × $0.20/GB direct = $100.00 | $2,000 model cost + 20% = $2,400.00 | $2,600.00 |
Note what these rows show: browser time is 2% or less of every total. The residential-vs-direct proxy decision swings the high-volume case by $2,400 per month on identical work.
How BU Pricing Compares
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Tool | Pricing Model | Starting Price | Best For |
|---|---|---|---|
| BU | Usage-based, prepaid credits | Credits from $5; browsers $0.02/hour | Browser automation where you want to pay only for runtime and traffic actually consumed |
| Dify | Open source + subscription | Sandbox $0 (200 message credits, 1 workspace, 1 member, 5 apps, 50MB knowledge storage); Professional $59/month per workspace | Teams wanting a predictable per-workspace subscription with a defined credit allowance |
| CrewAI | Freemium | Free tier 50 executions/month; Enterprise custom | Multi-agent orchestration where work divides cleanly into discrete executions |
| Flowise | Apache 2.0 source; Cloud sunset announced | Homepage displays Free $0/month, Starter $35/month and Pro $65/month; Cloud availability unverified | Teams prepared to maintain a fork and pay infrastructure/model costs separately |
The structural difference is what each price covers. BU meters runtime and traffic; Dify’s workspace subscription includes an allowance, and CrewAI’s free plan caps executions. Flowise’s announced sunset makes a maintained fork and operating costs the relevant comparison; hosted availability is unverified. Neither idle-time billing nor a subscription allowance establishes which option is cheaper or caps the complete bill. Model your measured workload and excluded costs.
Our recommendation: run a two-week measurement on the free tier, record GB-per-task and tokens-per-task on your actual targets, and decide the proxy setting deliberately before you commit meaningful credit. The default is the expensive one.