Get started

How it works

1
Import what your AI already knows

Paste a prompt into ChatGPT, Claude, or Gemini and it will export your preferences as fraglets. Instant library.

2
Connect via MCP

Add one JSON snippet to any MCP-compatible client. Your agent can then read, create, and manage your fraglets.

3
Your agent shares what's relevant

When a service needs context about you, the picker selects the right fraglets at the right level of detail.

4
Browse and build

Discover fraglets other people have shared. Associate with ones that fit, or adapt them to make your own version.

Why not just use your AI's memory?

Your AI probably already knows you well. The problem is that knowledge is locked inside one provider, buried in chat history, and impossible to share. Fraglets take what your AI has learned and make it portable, structured, and yours to control. Share specific preferences with specific services. Revoke access when you want. Use the same profile across providers. Your AI's memory is the starting point. Fraglets are what make it useful beyond a single conversation.

What is a fraglet?

A structured preference profile in a specific domain: music, food, travel, property, anything. Human-readable text plus a vector embedding, so it works for people and machines. Portable across services. Shareable by choice. Read the full concept at fraglet.org

Weeknight Sichuan loyalist
You eat out two or three times a week and default to Sichuan or Cantonese when choosing alone. You avoid large groups and prefer counter seating or small tables in places that don't take bookings.
food sichuan cantonese no-booking counter-seating weeknight

Shape how agents see you

Compose jackets — personas built from any combination of characteristics. Your own fraglets, ones you've picked up from others, adapted to fit. A date-night jacket might mix your food tastes with a cocktail fraglet you discovered. A work-travel jacket focuses on your hotel and solo-dining preferences.

Try on someone else's published jacket for a night, or adapt it to make it yours. A food critic's picks, a friend's going-out persona, an editorial "72 hours in Edinburgh" — jackets turn fraglets from a preference store into a creative medium.

Date night
3 fraglets across food, music, and travel. Intimate venues, good cocktails, interesting over expensive.
food music travel

Built on MCP

Fraglets use the Model Context Protocol. Any MCP-compatible agent — Claude, OpenClaw, or whatever comes next — can read, create, and manage your profiles. No special integration required.

🤖
Any MCP client

If your agent speaks MCP, it already works with fraglets.

🍽
MostMaker

London restaurant and live music recommendations, matched to your food and music fraglets. mostmaker.app

Fraglets describe how you choose, not who you are. No names, no locations, no personal data. Your agent shares only what's relevant, and only when you ask.

Get started