The SAP news landed in my inbox this morning: they led a round that values n8n at $5.2 billion. Five billion dollars for what is, structurally, a very pretty wrapper around OAuth tokens and webhook configs. I sat with that number for a while before figuring out why it bugged me.

n8n is genuinely useful. I’ve used it. Anyone who has tried to wire Salesforce to Slack to Notion knows that the plumbing is the actual work, and n8n makes that plumbing visible and editable. So the valuation is not crazy on its own merits. But it is strange that it’s happening in 2026, the same year browser agents have started doing the thing n8n was built to avoid.

What you actually buy when you buy n8n

A node. That’s the unit. Each node is a connector to a service, and each connector requires its own auth scope, its own rate-limit etiquette, its own schema knowledge, and its own breakage window when the upstream API ships a v2.

So when you build a five-step workflow, you are also signing five OAuth consent screens, granting five sets of permissions you will probably never audit again, storing five credentials in a vault somewhere, and trusting that the vault doesn’t get popped by a researcher with a Shodan query. The vendor manages this for you, which is the whole product. The “automation” is mostly the credential vault and the retry logic.

This is the integration tax. And SAP just paid $5.2 billion to be the landlord.

The browser already did the integrations

Every service in your n8n flow already has a working UI. You’re logged into it right now. Your session cookie is sitting in Chrome with exactly the scopes you need, scoped to your account, expiring on a schedule you don’t have to think about.

So a browser agent that runs in your Chrome doesn’t need to chain APIs because the Salesforce tab and the Notion tab are already open, already logged in, and already authorized to do whatever your account can do. The Salesforce CRM doesn’t need a Salesforce node. The Notion database doesn’t need a Notion connector. Your already-authenticated session is the integration, and it covers every site you have a login for, including the ten thousand SaaS tools nobody bothered to build a node for.

This is the Browser Agents Are the New API argument, but with a price tag attached. SAP just put a number on the thing browser agents are quietly disintermediating.

Every node is a place data leaks

Workflow tools paper over a hard security problem. To stitch ten services together, you have to hold persistent credentials for all ten in one place. And that place is, by construction, the juiciest target in your stack. When self-hosted n8n instances get popped on Shodan (and they do, regularly — go look if you don’t believe me), the attacker gets a master key ring to every connected service in a single shot.

But a browser agent running locally in your Chrome holds nothing it didn’t already have. It uses the session you’re already using. So if your laptop is compromised, your sessions were already compromised, and adding the agent didn’t meaningfully widen the surface. The threat model is dramatically smaller, because there is no central credential store to breach, no cloud worker logging tokens, and no third party sitting in the middle holding a copy of your data while it pretends to be a Zapier alternative.

I don’t think most of the n8n customer base has done this math yet. They will eventually. When they do, an extension like Dassi that runs inside your already-logged-in tabs will start looking less like a curiosity and more like the only thing that makes any damn sense for sensitive work.

The MCP detour

Some people will argue MCP solves this. It does not. MCP is integration plumbing with a fresher logo, covered in 40 MCP Tools and It Still Can’t Fill Out a Form.

Why workflow tools still survive

n8n is not going to zero. And cron jobs and server-to-server pipelines need exactly what n8n is selling. If you’re moving rows between databases at 3 a.m. with no human in the loop, you want a server with API keys, not a Chrome window. That use case is real and it isn’t going anywhere.

But that use case is also nowhere near $5.2 billion of total addressable market. The big number SAP put on the table is a bet that workflow automation also includes the things humans do during the day at their desks, the form-filling and the data-pulling and the cross-tab reconciliation. And that’s the work browser agents are eating from the other end of the plate, without asking anyone for an OAuth consent screen.

A closing observation

The pattern in enterprise software lately is that whoever paid the most for a category usually showed up right after the category started shrinking. So SAP paid eight figures per node, by my back-of-envelope math. Dassi paid zero per node, because there are no nodes. I know which of those numbers I would rather defend at the next board meeting.