Inspiration
Memory books are built from many small pieces: photos, stories, memories, events, and messages submitted by different people. Collecting those contributions is relatively easy. Turning them into a coherent and meaningful book is much harder.
MemoryBook explores a different approach. Instead of using AI only to generate text or answer questions, the application exposes meaningful book operations through WebMCP. This allows an AI agent to understand the contents of the book and actively help organize and build it.
The goal is to make AI a useful collaborator inside a real web application while keeping the user in control.
What it does
MemoryBook is a collaborative digital memory book application where people can submit memories and photos and an organizer can turn them into a finished book.
Through WebMCP, an AI agent can interact with the application using structured tools.
The current integration allows the AI to:
- retrieve the current MemoryBook status and context
- list submitted contributions
- analyze which important school-year events are missing
- propose a thematic spread based on the submitted memories
- build that thematic spread directly in the MemoryBook
- reorder pages in the book
This creates an end-to-end workflow in which the AI can first understand the collected material, reason about it, propose an editorial decision, and then make an actual change to the book when requested.
How we built it
MemoryBook is a full-stack web application built with React, TypeScript and Vite on the frontend, with a Node.js/Express backend and PostgreSQL for persistent data.
The visual editor supports editable text, drawing, images, layers, undo/redo, page navigation, saving and page previews.
For the WebMCP integration, I designed tools around meaningful user intentions instead of low-level interface controls.
These include tools for retrieving book context and status, listing contributions, checking event coverage, building a thematic spread, and reordering pages.
The AI therefore does not need to simulate mouse clicks or manipulate individual interface elements. It can discover structured capabilities exposed by the website and use them to work with the same MemoryBook data as the normal user interface.
Challenges we ran into
The main challenge was connecting AI actions to a real, persistent editor rather than creating a scripted demonstration.
WebMCP operations had to work with the same pages, contribution data, saved state and previews used by the normal application.
Another challenge was choosing the right level of abstraction for the tools. Low-level controls would make the AI responsible for operating the UI. Higher-level WebMCP tools allow it to work with concepts that actually matter to the user: contributions, missing events, thematic spreads and page order.
Reliable saving and synchronization were also important because an AI-created page must behave like any page created or edited manually.
Accomplishments that we're proud of
The most important accomplishment is a working end-to-end WebMCP workflow.
The AI can inspect real submitted memories, analyze their coverage, propose a thematic spread, and then build that spread directly in the MemoryBook.
The resulting content is not just an AI response in a chat window. It becomes part of the actual editable book.
The project also demonstrates that WebMCP can support both read-only analytical operations and actions that modify application state, while allowing the user to decide when an operation should be executed.
What we learned
Building MemoryBook changed how I think about AI integration in web applications.
An embedded chatbot can answer questions about an application. WebMCP enables something different: the website can expose what it is capable of doing, and an AI agent can use those capabilities as tools.
The most useful model is not necessarily AI replacing the interface.
A person may prefer a visual editor for creative decisions. An AI agent may be better at reviewing many contributions, identifying missing themes and organizing information. Both can work with the same underlying application and book.
What's next for MemoryBook
MemoryBook is currently a working prototype focused on demonstrating the WebMCP workflow.
The larger direction is a complete collaborative memory-book platform where groups can collect memories and photos, organize them with AI assistance, visually edit the result, and produce a finished digital or printable book.
Future development can expand the collaboration workflow, improve AI-assisted layout and organization, and add production-ready publishing and export capabilities.
WebMCP can make this process substantially easier by allowing AI agents to handle organizational and repetitive editorial work while keeping people in control of the final book.
Log in or sign up for Devpost to join the conversation.