Project Name
MemoryLane
Project Description
A personal photo library powered by Directus that you can query in natural language through MCP.
We all have thousands of photos scattered across devices. While our phones already have face recognition built-in, this project demonstrates how Directus + MCP can power conversational data access for any use case — from personal memories to professional workflows like wedding photographers managing thousands of tagged shots in Lightroom. (Imagine a website with a firstname input that allows to retrieve all photo of a single person, cross group, and more…)
The real insight: MCP isn’t just a dev tool. It’s a way to give end users direct access to their data without code, queries, or complex interfaces.
Challenge Entered
Project Challenge
Team Details
MrCasquette (solo) — Vincent, freelance fullstack developer from France
Technical Details
-
Backend: Directus with native MCP server enabled
-
Data: Personal photos with people tags preserved in XMP metadata (
metadata.xmp.subject) -
Client: Claude Desktop connected via MCP
-
Custom system prompt to guide photo-related queries
No custom schema — just the native directus_files collection with XMP metadata.
Starting Point Details
I’ll be honest: I rushed this project in one evening due to time constraints. But I really wanted to participate — I love this community and want to get more involved.
Started from scratch with:
-
Fresh Directus instance with MCP enabled
-
Personal family photos exported with XMP tags
-
Custom system prompt for the photo library context
Demo Video
What’s Next
This is a proof of concept. The real vision is a universal template where end users can:
-
Describe what they want to manage
-
Have MCP create the schema
-
Populate data through conversation
-
Query everything in natural language
No Directus UI. No code. Just chat. That’s my next project.
