Browser Use Cloud pricing, and what self-hosting costs instead
Their published rates, read on the date below, and the same work priced on hardware you own. The difference is not a discount, it is a multiplier applied to tokens you were going to buy anyway.
What Browser Use Cloud charges
Browser Use Cloud bills a monthly plan for concurrency, then usage on top. These are the rates as published:
| Item | Rate | What that means |
|---|---|---|
| Agent tokens | 1.2× the model provider's own rates | Charged on top of what the tokens would cost you at the provider. |
| Agent tokens, bring your own key | provider rates, plus a 0.2× orchestration fee | You pay OpenAI or Anthropic directly, and the platform fee arrives separately. |
| Browser session | $0.02 per hour | Billed per minute for the time a browser is open. |
| Managed proxy | $5 per GB, $4 per GB on Scaleup | Their residential proxy, on by default. |
| Egress without their proxy | $0.20 per GB | Charged even when you bring your own proxy or use none. |
Source: browser-use.com/pricing
Plans buy concurrency and arrive as credits: Free at $0 a month for 10 concurrent sessions, Dev at $29 a month for 25 concurrent sessions, Business at $299 a month for 200 concurrent sessions, Scaleup at $999 a month for 500 concurrent sessions. Annual plans get two months free, with credits granted up front.
What the same work costs self-hosted
Self-hosting removes the platform from the bill entirely. What is left is the same tokens at the provider's own price, plus a machine and the electricity it draws.
| Item | Rate | What that means |
|---|---|---|
| Agent tokens | you pay the provider directly | You hold the API key. Nothing is added on top. |
| Browser session | nothing | The browser runs on your machine, for as long as you like. |
| Egress | your own connection | Requests leave from your address. There is no proxy layer and no per-gigabyte meter. |
| Hardware | one-off | It was built and benchmarked on a Raspberry Pi 5, which is the whole of the fixed cost. |
What you give up for that is a managed residential proxy, session recordings, workspaces and somebody else's uptime. The full comparison covers all of that dimension by dimension.
The multiplier is the whole story
Both of their token routes land in the same place, and it is the number worth carrying away.
Managed tokens are billed at 1.2× the provider's own rates. Bring your own key instead and you pay the provider directly, plus a 0.2× orchestration fee. Either way the platform's share is 20% of a bill you were always going to pay, and self-hosting is the same tokens at 1×.
Applied to something measured: the fastest complete run on our benchmark cost $0.24 in tokens on gpt-5.6-terra. The same tokens through the platform would be $0.29, before session time at $0.02 per hour and any egress. On one run that is pennies. It is a percentage, so it scales exactly as your usage does.