Iv been reading some threads but i coudnt find now one talk about this yet. Recently I’ve noticed a significant increase in the usage costs of the n8n AI Assistant, and I’m honestly quite concerned about it.
I’m a non-technical user with no coding experience, so the AI Assistant is one of the main reasons I’m able to use n8n and build workflows in the first place. I rely heavily on it to understand how things work, troubleshoot issues, and create automations that I wouldn’t be able to build on my own.
I still think the N8N is an amazing product, that’s why I did my first pay 15 days ago after creating my scrapper but im not sure if its just me of anyone else found this issue very recently. I feel the costs for AI usage went up 20x and i can easily say, the scrapper i build during the “Free 15 days” i couldn’t even do 10% on it now with the 800 credits they give us to test it out.
A big part of what you’re seeing is how the assistant is billed: credits meter tokens, and every message you send re-sends the whole conversation so far plus your workflow context. That makes a long-running chat cost many multiples of the same question asked in a fresh one, and it compounds as the chat grows.
Things that helped me: start a new assistant chat per task instead of keeping one open forever, ask for one node at a time rather than “build the whole workflow”, and paste only the failing node’s output rather than the full workflow JSON each time. If your plan shows an AI usage breakdown, check which model ran and messages-per-day there, since a heavier model moves the credit rate on its own.
Also worth knowing before you judge the jump: trial credits are priced against the bundled model, so a fixed pack of them was never going to equal the same number of messages you had during the free period. If the volume is real work you’ll keep doing, the cheaper long-run route is the community edition on your own small server with your own model key, where you pay provider token rates directly instead of credits - the tradeoff is that you maintain the box yourself.
Iv been trying the gather info on how to safe tokens, for me the small server its not an option for now. Iv been using the N8N ai for a month and felt that since Octobers 1º week there was a change in the price per usage. Maybe its just me, but i relly heavily on the AI assistant as i have no Dev or coding knowledge
You’re not imagining it. On October 1, n8n changed how Assistant credits work: the Assistant moved to a newer model, every credit now has the same fixed value across paid plans (because credits can now be bought), and each plan’s monthly allowance was reset to match. Details in the announcement: Assistant credits top up - now available
What that means for you:
The 800 trial credits were a one-time allowance. Paid plans get a monthly allowance that resets on the 1st: 1,600 on Starter, 4,000 or 9,600 on Pro depending on the tier.
n8n’s own rough guide: a quick question uses 10–20 credits, a simple workflow or fix from one prompt 50–100, and a bigger build or long debugging session 300–600. A scraper built through lots of back-and-forth can easily use a few hundred.
If you run out mid-month, the instance owner can now top up from the Assistant credits tab in the Cloud admin dashboard. Bought credits carry over and last 12 months.
Running your workflows never uses Assistant credits. Only chatting with the Assistant does.
+1 to HomelabAgent’s tips. n8n’s post says the same thing: longer conversations use more credits and specific prompts use fewer, so a fresh chat per task with one detailed prompt goes furthest.
On the DM: new forum accounts often can’t send private messages right away. Posting your n8n version here should be enough for riqwan to start.