Browser Use Cloud charges $0.02 per browser-hour, and our first reaction to that number was that nobody can judge it without knowing how long their agents run. That’s a very developer-shaped problem to have, and a decent summary of who Browser Use is for.
The core of it is a free, open-source Python library (MIT license, around 115k stars on GitHub) that you install on your own machine, point at whichever LLM you like, a local one through Ollama included, and drive from your own code. It can reuse your Chrome profile. The Cloud side hosts agents in anti-bot browsers, with a $15 one-time credit to start and proxy traffic at $5 per GB on top of the hourly rate and your LLM tokens.
What the library leaves to you
What Browser Use hands you is an agent loop and control of a browser, and after that every decision is yours: which tasks run, where results end up, what happens when a step fails and nobody’s watching. If you’re building an agent into a product you ship, those are the decisions you’d want to make yourself.
We go back and forth on who else it suits. Plenty of people who aren’t developers can follow a Python tutorial well enough to get a first script running, and some of them clearly enjoy it, so we don’t think “non-developers should stay away” is right. What we’re less sure about is the second month, when the script needs a new login step or a change in what it does, and the person who wrote it has moved on to other things.
That weekly job, in dassi
Every Monday, say, you pull last week’s orders from a supplier portal, check them against a Google Sheet, and email the two or three suppliers whose deliveries slipped (we keep using supplier portals as the example because they’re the worst-behaved sites we know, full of session timeouts and forms that clear themselves if you wander off for a coffee). As a Browser Use project, that’s a script written from scratch, since there’s no recorder, along with whatever login handling it needs and some way of starting it every Monday morning.
In dassi you’d describe it roughly the way you’d explain it to whoever covers for you on holiday, or just do it once while dassi follows along, and it becomes a workflow you can run on demand or on a schedule. Scheduled runs happen in your own browser, which needs to be open at the time; a laptop asleep in a bag won’t manage it. Sending the supplier emails can be flagged so each one waits for you, and if the portal shows something dassi can’t make sense of, it asks you rather than guessing.
Logins and 2FA
Beyond borrowing your Chrome profile for sites you’re already signed into, logins and 2FA are extra work in a Browser Use script.
dassi works in the Chrome, Edge or Brave you already use, on the accounts already signed in there. When a session has lapsed, it signs in with whatever Chrome’s password manager autofills; the password stays in your password manager, and you never paste it into a chat or into dassi’s settings. The 2FA code is on your phone, and dassi waits.
When to choose Browser Use
If you write Python and want an agent inside your own product or pipeline, use Browser Use. You get every line under a permissive license, and the whole thing can run locally with a model through Ollama.
When to choose dassi
It’s $10 a month with your own key, or $25 to $300 a month on a plan with credits. The agent runs locally, and what it reads goes to the model provider you picked, via dassi’s relay if you’re on included credits. It’s built for the Monday job above and the dozen like it that sit on a sticky note for months, because each one only takes twenty minutes and twenty minutes never quite justifies an afternoon of Python.