SAP put a $5.2 billion valuation on n8n this week, and the tech feeds lit up the way they always do when a number that big lands on an open-source-flavored tool. There were two threads trending side by side on Hacker News: one about the funding, one a tutorial on building n8n AI automation workflows. I read both back to back, and the contrast stuck with me.

Because the funding story is about how huge workflow automation has gotten. And the tutorial is about wiring up your fourth OAuth connection so a node can finally read your Gmail.

The thing $5.2B doesn’t buy you

n8n is genuinely good software. I’m not here to dunk on it. If you need a scheduled pipeline that pulls rows from Postgres, runs them through a model, and drops the output into a Slack channel at 6am with nobody watching, it’s hard to beat. That’s the job it was built for and it does it well.

But every one of those nodes is an integration. The Gmail node needs Google OAuth. The Notion node needs a Notion token scoped to the right workspace. The CRM node needs an API key, and if your CRM is some niche vertical SaaS your industry forced on you, there might be no node at all, at which point you’re staring at the generic HTTP Request node trying to reverse-engineer an undocumented endpoint from your browser’s network tab. The $5.2 billion doesn’t change that math. It scales it.

And the math is worse than it looks at the start, because integrations aren’t a one-time cost. APIs deprecate. OAuth scopes get tightened. A vendor decides their v2 API now requires a paid tier, and your perfectly good workflow breaks on a Tuesday for reasons that have nothing to do with anything you did.

You already authorized the app. You did it weeks ago.

Here’s a fact that I think gets lost in all the connector talk: you are already logged into every app you care about. Right now. In tabs that are open. Your Gmail session is live. Your CRM is one click away. Your bank, your analytics dashboard, the internal admin tool with the ugly table — all authenticated, all sitting there.

An integration is, in a sense, a way of re-asking for permission you already granted. n8n can’t see your logged-in Gmail tab, so it has to go around the side of the building and knock on Google’s API door with its own credentials. That door is guarded by OAuth, by scopes, by consent screens, and increasingly by security teams who do not want a third-party automation platform holding a long-lived token to the company inbox.

A browser agent goes through the front door you’re already standing in. It works inside the tab. It reads what’s rendered, clicks what you’d click, types what you’d type. There is no connector because there is nothing to connect — the authentication already happened when you signed in this morning. This is roughly the whole idea behind dassi, which lives in a Chrome side panel and acts on the page you’re looking at.

What this actually looks like

Say you want to pull every invoice over $500 from a vendor portal that, of course, has no API. In n8n, this is a non-starter or a brittle scraping hack. In a browser agent, you open the portal you log into every month and ask it to read the table and give you the rows. Done.

Or you want to draft replies to a batch of support emails sitting in a shared inbox. n8n’s Gmail node can do it, after you set up the OAuth app, configure the scopes, and accept that an automation server now has standing access to that mailbox. The browser agent drafts them in the Gmail tab that’s already open under the account that already has access. No new trust relationship gets created.

The difference isn’t capability. Both can read an email and write a reply. The difference is everything that has to be true before the first useful thing happens.

The plumbing is the product, until it isn’t

What strikes me about the $5.2B number is what it says about how much of automation is still plumbing. A big chunk of n8n’s value, like Zapier’s before it, is the catalog of pre-built connectors. Thousands of them. That catalog is a moat and a maintenance burden at once, a standing army of little OAuth handshakes someone has to keep alive as hundreds of vendors change their APIs on their own schedules.

Browser agents sidestep the catalog. There’s no list of supported apps because the supported app is “whatever renders in a tab.” That covers the long tail of internal tools and weird vertical SaaS no integration platform will bother building a node for, which is precisely where most people lose their afternoons.

I don’t think this kills n8n. The two approaches genuinely solve different problems, and the unattended-server-pipeline job is real and won’t go to a browser tab. But I do think the framing of “automation = integrations” is going to age strangely. We’ve already written more about why browser agents fill the API gap and how the no-API case keeps showing up, and the n8n valuation just put a very large dollar figure on the part of the problem that browser-native execution makes disappear.

So congratulations to n8n, genuinely. Five billion is a lot of OAuth screens. Just maybe don’t set up the fourth one before checking whether the tab’s already open.