On September 17, 2026, n8n shipped v2.36 and removed the single biggest friction point in AI-powered automation: the API key dance. If you've ever tried to wire up a new AI model inside a workflow only to detour through three signup pages, a credit card form, and a rate-limit dashboard, you know exactly what they killed.

What Gateway credits actually are

The concept is dead simple. When you add a supported AI node to an n8n Cloud workflow — say Anthropic, Google Gemini, or OpenAI — you now get a second credential option: Gateway credits. Pick that instead of pasting in your own API key, and the request runs through n8n's prepaid balance. No new account. No new billing relationship. No API key to rotate or leak.

Six model providers and five tool services are live today. The list reads like a who's-who of the AI stack: Anthropic, Google Gemini, OpenAI, Mistral, and others on the model side, plus tool services for things like web search and document processing. If you later decide you want your own direct relationship with a provider, bring-your-own-key works exactly like before.

Why this matters more than it sounds

The obvious benefit is convenience. You open n8n, drop in a node, and it runs. That's the pitch, and it's real. But the second-order effect is where it gets interesting.

Small teams and solo builders rarely experiment with unfamiliar AI models because the activation energy is too high. You need to evaluate pricing, create an account, add a payment method, generate credentials, and manage yet another vendor invoice. Most people just default to whatever they already have keys for, even if a different model would be better for the task.

Gateway credits collapse that decision. Want to see if Gemini handles your summarization workflow better than Claude? Toggle the credential, run it, compare the output. No signup, no new billing line item. The switching cost drops from twenty minutes of account management to a single dropdown.

n8n says they're pricing Gateway credits at each provider's published list price — no convenience markup. Your balance shows in real currency, and the admin panel breaks down usage by workflow and by service. They're not trying to make margin on the passthrough; they're removing a barrier so more people build more workflows on their platform.

The privacy tradeoff

There's one detail worth noting. When you use Gateway credits, the provider receives your request content — the prompt, the document, whatever you're sending — but not your identity or n8n account details. n8n acts as the intermediary.

If you're processing sensitive client data or anything regulated, that's a different calculus than using your own direct API key where you control the data processing agreement. For personal projects and small-team experimentation, it's probably fine. For a healthcare workflow or anything touching PII, you'd want to think carefully about whether the convenience is worth the added indirection.

What this signals about the automation market

This move puts n8n in a position that Zapier and Make haven't tried yet. Zapier has "AI by Zapier" with its own model-tier pricing (which, as of June 2026, can get expensive fast). Make has integrations but still requires you to bring your own keys. n8n is saying: we'll be the single pane of glass for your entire AI workflow, and we'll make the billing boring on purpose.

It also hints at where self-hosted automation is heading. n8n's self-hosted version doesn't get Gateway credits — that's a Cloud-only feature. The bet is that enough builders will pay for the hosted version to skip credential management, while tinkerers who want full control can still self-host and BYOK.

The bottom line

If you're on n8n Cloud and building AI workflows, Gateway credits are worth trying today. The activation energy for experimenting with new models just dropped to near zero, and the pricing is transparent enough that you're not going to get surprised.

If you're on Zapier or Make and feeling the AI tax creep up, this might be the nudge to prototype your next workflow in n8n. Not because it's cheaper — the per-call pricing is roughly the same — but because the path from "I have an idea" to "it's running" is now shorter than it's ever been.

The era of managing six API keys for one automation stack is ending. n8n just made the first move.