What is Browser Use?
Browser Use is an agent framework for turning a task into browser actions. The framework handles the agent workflow; Remote Browser supplies the Chromium runtime that loads pages and holds active browsing state. Keeping those responsibilities separate lets you choose how the agent runs without placing a browser process on the same worker.
The official remote-browser configuration documents attaching through a CDP URL. Browser Use's own cloud service is a separate provisioning option; credentials for that service are different from your Remote Browser endpoint.
Why pair Browser Use with a remote browser?
Consider this setup when your agent worker needs an independently managed browser, authorized account reuse, or a human-visible session for troubleshooting. A remote runtime also gives you a clear place to configure the browser's network route and resource budget.
A useful acceptance test is an authenticated, read-only task: open a known dashboard, verify the account, read one value, and save appropriate evidence. Define success at the website level rather than treating a completed model response as proof that the browser task succeeded.
Quick start
Create a dedicated session using the Remote Browser quick start. Put its authenticated connection URL in REMOTE_BROWSER_CDP_URL using your environment or secret manager. In a project with a pinned browser-use dependency, configure the browser as follows:
import os
from browser_use import Browser
browser = Browser(cdp_url=os.environ["REMOTE_BROWSER_CDP_URL"])Pass this browser object to your Agent together with your task and supported model configuration. This snippet configures attachment; it does not configure an LLM or execute a complete task. Model credentials and the browser endpoint must both be available to the agent process.
Connection methods and ownership
CDP is the client-to-browser connection. Your application or provider provisions the session before attachment and releases it afterward. A browser API is the lifecycle interface around that runtime; it is not another name for CDP.
Set the profile and browser-side proxy route when allocating the browser. Define whether one worker owns a session, how a reconnect locates it, and how a cancelled task triggers cleanup. Avoid sharing a writable account profile between unrelated concurrent tasks.
Use cases
- Cloud workers that need a browser without maintaining a local Chrome installation.
- Authorized dashboard research that reuses a saved login.
- Long workflows with a human available for authentication renewal.
- Parallel jobs with separate sessions, account boundaries, and resource limits.
For each case, measure failed attempts as well as successful completion. A persistent profile cannot prevent a website from revoking its login.
FAQ
Does using CDP require Browser Use Cloud?
The upstream configuration supports a supplied CDP endpoint. Choose the provisioning service separately, and keep each service's credentials distinct.
Does a saved profile keep the agent running?
A profile saves supported browser data. It does not keep an agent worker alive or restore an unfinished decision loop. Plan checkpoints and browser cleanup in your application.
How should I compare local and remote runs?
Keep the task, model, prompt, and success criteria fixed. Record browser and worker regions, account state, proxy route, failures, and cleanup. The timing methodology below measures the browser connection path, not the complete Browser Use agent loop.