Pro · Off by default

It can't change your trip. That's the point.

RPlan's assistant reads your itinerary and suggests things you might have missed. It has no ability to add, edit or delete anything. Every idea arrives as a proposal with an Apply button, and until you press it nothing about your trip has moved.

Reads your trip Writes nothing Off until you enable it
Assistant
Kyoto, autumn · Example trip
Anything I should book before we go?
Two things stand out. Fushimi Inari on your Saturday needs no booking, but the kaiseki dinner on the 17th may need a reservation. You also have no JR Pass on the checklist and four intercity legs planned.
Proposed · Checklist
Book the kaiseki dinner for the 17th
Not added yet · you choose
Apply Dismiss
What it does

It already knows the trip you're asking about.

The assistant is a panel on the trip page, not a separate chatbot you have to brief first. It can see that trip's dates, days, plans, times, checklist and packing list, which is why it stops suggesting things you've already got.

Ask

"What am I forgetting to pack?" "Anything I should book ahead?" "Is Tuesday too full?" Starter questions are one tap, or type your own.

Suggest

A Suggest button on Plans, Checklist and Packing asks for ideas scoped to exactly that list, rather than a general answer about your destination.

Proposals

Plans, checklist items and packing items come back as separate cards. Apply one, dismiss another, ignore the rest. They're independent.

You place it

When you apply a proposed plan, you choose the day it lands on. The assistant never picks dates, places or prices for you.

Where the limits are

The parts most AI features don't mention.

Written plainly because the honest version is the selling point here.

It has no write access at all

Applying a proposal goes through the same validated endpoints your own typing does. There is no path by which the model edits your trip directly. It is not a policy, it is an architecture.

It can be confidently wrong

Opening hours, prices and whether somewhere needs booking are exactly the facts a language model invents most convincingly. Treat every suggestion as a prompt to go and check, not as a source.

Travel times don't come from it

The times on your gap rails come from Google's Routes API, not from the assistant. A model guessing a drive time is wrong in exactly the places that matter: a bridge, a ferry, rush hour.

It's off until you turn it on

Per account, in your settings, and Pro-only when enabled. Nothing about your trip is sent anywhere until you open the panel and ask something.

There are rate limits

So the cost stays predictable for both of us. In normal planning you won't meet them; if you do, they reset.

You can ignore it entirely

Everything else in RPlan works exactly the same with the assistant switched off, and always will. It's a convenience, not the product.