Everway.
From scattered bookings to one useful trip.
View project Adrian built Everway to bring travel details together without asking travellers to hand over their inbox. The product also puts our own AI-assisted development approach to work on a complex everyday problem.
The challenge
A single trip can involve confirmations, tickets and attachments from many providers. Finding the right detail at the right moment becomes a task of its own, especially when travelling with other people.
The approach
The starting point is an action people already understand: forwarding an email. Each trip has a private address, giving the product a clear boundary around what it receives. Our in-house prompts and agent memories support the development workflow behind that experience, separately from the AI features travellers use inside the app.
Useful information, with a source
Everway extracts booking details from forwarded emails and attachments, including PDFs, screenshots and calendar invites. It organises them by day, flags uncertainty and links items back to their source.
Built around the journey
Mapped stops, shared group itineraries and offline access help keep the plan useful during travel. Personalised preparation checklists and reminders extend the experience beyond storing confirmations.
Assistants that know the trip
An MCP integration lets compatible AI assistants work with real itinerary information, including checking what remains to be booked and adding trip items.
OUR APPS.
OUR PRACTICE.
Everway makes the distinction between building with AI and building an AI product tangible. Agent prompts and memories support our delivery process; the application’s parsing and assistant connections serve travellers. Keeping those responsibilities clear helps us reason about the experience from end to end.
Across these apps, our in-house prompts give agents direction and our memories carry useful context forward. We use our own products to refine that approach, with the aim of developing production applications quickly and precisely.
- 01 / DIRECTION
Give the work a clear brief.
Start with the problem and the intended behaviour. A focused prompt gives an agent something concrete to work towards.
- 02 / CONTEXT
Carry the learning forward.
Agent memories keep useful context available between iterations, so each task can build on the thinking that came before.
- 03 / APPLICATION
Bring the workflow to your business.
Apply the same approach to a new product or an existing process. Production quality remains a design and engineering responsibility.
THE WORK,
UP CLOSE.
- 01Private email addresses for individual trips
- 02AI extraction with uncertainty flags and source links
- 03Day-by-day itineraries, maps and group planning
- 04Offline access and travel preparation tools
- 05Connections to compatible AI assistants
A live travel product that brings fragmented information into one plan. Building it gives us a place to keep refining the prompts, context and product judgement we bring to other businesses.
BUILT BY ADRIAN CIASCHETTI. COVER: SCREENSHOT OF THE PUBLIC HOMEPAGE.