There’s a lot of talk right now about restaurant booking bots. People are posting their own n8n workflows, screenshots of a chat window, “look what I built.”
But how hard is it, actually, to build one that works? Would it handle a real, messy request — someone wanting to book a table and check if there’s a vegetarian option, in the same message?
I built one and put it online so you can try it yourself.
On one page: the table layout, the menu, restaurant details, and a chat window to talk to the bot directly. Book a table, ask about the menu, throw a complex multi-part request at it and see what it does. You can also switch to the owner’s side — reserve or cancel tables yourself — and watch the bot see the exact same state you just changed.
Free, no registration: try it here.
It runs on n8n with an OpenAI agent at the core, tracking each request as its own ticket so it can handle a reservation, a menu question, and a staff note at the same time without losing track of any of them. More on how it’s built in the case study.
Three nearby posts worth opening next.

May 21, 2026
At 5-10 minutes per invoice, 300 invoices a month is 25-50 hours of manual entry. The part most people skip is building the GL mapping table.

May 18, 2026
A student built an n8n invoice controller where a cheap model as the first filter turned out to be the most valuable node in the workflow.

May 6, 2026
PDF invoice extraction is the easy part. The hard part is what happens when the invoice doesn't look like an invoice.
If you have a manual workflow between tools, I can help map the logic, design the system, and automate it in a way your team can actually use.
Hire Me on Upwork