SAP reportedly led a round valuing n8n at $5.2 billion, and the number ricocheted around Hacker News all morning. n8n is a workflow-automation tool. You drag nodes onto a canvas, wire a trigger to an action, and the thing runs without you babysitting it. Genuinely good product, and the valuation isn’t crazy once you notice how many teams have quietly standardized on it for the glue work nobody wants to write by hand.

But one detail kept nagging at me while I scrolled the comments.

What the valuation is actually pricing

Every node on that canvas needs a connection. Before n8n can touch your Gmail, your Notion, your Stripe account, or the internal CRM your company built in 2019, you have to authorize it first. An API key here. An OAuth grant there. A service account for the one tool that still doesn’t do OAuth properly and makes you paste a token into a field that warns you, in red, never to share this token.

The canvas is elegant. The plumbing under it is a swamp.

So when SAP writes a check that size, a big chunk of what they’re buying is the integration library. Hundreds of pre-built connectors, each one a small treaty negotiated with somebody else’s API. That’s the moat. It’s also the tax, because the moment a service you use isn’t in the catalog, or changes its auth flow, or sits behind a login with no public API at all, the elegant canvas just stares back at you.

The browser already did the authorizing

Think about how you actually use those same services. You opened a tab. You logged in, maybe two months ago, and the session has been sitting there ever since. Gmail knows you. Stripe knows you. The CRM knows you. You didn’t mint an API key to read your own email this morning; you just clicked the tab and it was there.

A browser agent works inside that. It drives the page you’re already authenticated on, the same way you would, except faster and without complaining. There’s no connector to install because there’s no second system asking for permission. The permission already happened when you logged in, and it lives in the cookie jar, not in a settings panel full of scopes you half-understand and fully clicked “allow” on anyway.

This is the part workflow tools structurally can’t reach, and it’s worth sitting with for a second, because the gap isn’t a missing feature that n8n will ship next quarter. It’s the difference between a system that talks to services through their front-facing programmatic contracts and a system that simply operates the same interface a human operates, which means the long tail of weird internal dashboards and login-walled vendor portals stops being a special case and starts being Tuesday.

A small example

Say you need every invoice over $500 from a vendor portal that has no API. None. It’s a web app a human logs into, full stop.

In n8n, that’s where the flow dies. No connector exists, so you’re back to building a custom HTTP node, reverse-engineering their endpoints, and praying they don’t rotate a CSRF token on you.

A browser agent just opens the portal you’re logged into, reads the table, filters the rows, and hands you the list. Dassi does this in the Chrome side panel while you watch it happen. No token. No scope. No treaty.

So is n8n dead? No.

Obviously not. For backend-to-backend automation that runs at 3am while you sleep, with no browser open and no human anywhere near it, a workflow engine with real connectors is the right tool and a browser agent is the wrong one. That world is real and it’s big and SAP is right to want a piece of it.

But a lot of what people reach for n8n to do is not that. It’s reading a page, pulling some numbers, drafting a reply, moving data from a thing you’re looking at into another thing you’re looking at. That work doesn’t need a server farm or a connector catalog. It needs something that can see the tabs you already have open, which is exactly the seam that browser agents fill when there’s no API to wire into.

I’ve watched teams spend a full afternoon getting an OAuth app approved internally just to automate one report. That’s an absurd amount of ceremony, and half the time the pile of integrations still can’t fill out the form at the end of it.

The thing I keep coming back to is that the most valuable connector is the one you never have to build, because the login already happened. Dassi runs that way: you can add it from the Chrome Web Store, point it at a tab you’re signed into, and bring your own model key so the whole thing stays between you and your LLM. No $5.2 billion catalog required. Just the tab that was open the whole time.