browser-use drives a real Chromium browser from an LLM. Where that browser connects from decides what your agent sees: local search results, regional prices, which store page loads. This guide puts the agent behind a dedicated 4G/5G mobile proxy in about ten lines, then proves it worked.
You need four things: the proxy host, the HTTP port (8000 on RelayKit Multi Location and Single Location proxies), and the username and password. In the RelayKit dashboard, open My Proxies, expand a proxy card and click 🤖 Use with AI tools. Pick browser-use and the snippet below appears with your real values already filled in, plus a copy button. The rest of this guide uses placeholders.
You also need an LLM. The examples use ChatBrowserUse(), browser-use's bundled model; any chat model browser-use supports works the same way. Put the key it needs in a .env file.
The proxy goes on the Browser object as a ProxySettings, and the browser is handed to theAgent. Use the HTTP port here (see section 5 for why).
The task above is the verification. The IP the agent reports must match the IP shown on your proxy card in the dashboard. If it reports your own IP instead, the proxy settings didn't apply — check the host and port, and thatProxySettings is imported from browser_use.browser.
Tip: keep this one-line IP task as the first step of every new agent project. It takes a few seconds and removes the most common "why are my results wrong" question before it starts.
Chromium, which browser-use runs on, can send a username and password to an HTTP proxy but not to a SOCKS5 proxy. That's why the config above uses port 8000. If you'd rather use SOCKS5 (port 9000), turn on IP Whitelisting on the proxy card and enter the public IP of the machine running your agent. The proxy then accepts that machine without a password, and the config shrinks to:
The dashboard's Use with AI tools panel detects whitelist mode and shows the no-password version automatically. Remember that a whitelisted proxy rejects every other machine — turn whitelisting off before running from somewhere else.
The point of a mobile proxy for an agent is the place, not just the IP. A RelayKit Multi Location proxy sits in one US city and can be moved to another from the dashboard whenever you like. That makes an agent task like this meaningful: run it, switch the city in the dashboard, run it again, compare.
Typical uses: checking how a competitor's pricing or availability differs by region, confirming a geo-targeted campaign actually renders in the target city, local SEO checks, and QA of a site's own location-based behavior. For a complete multi-city runner that verifies the city before each pass and writes a CSV, see Run an AI research agent across multiple US cities.
Browser(proxy=...) is the browser you pass to the Agent, and that the server starts with http://.Dedicated 4G/5G proxies with unlimited data, HTTP and SOCKS5, and a Multi Location plan that switches between US cities from the dashboard. Pay with crypto; sign up with an email.