Mime

Inspiration

When forecasting the future of Agentic AI, we realized people won't be browsing anymore, they'll be asking AI to shop for them. But the moment AI tries to, it runs into systems that weren’t built for it, they were built for humans. Results break, data is messy, and merchants lose control of what is perceived by the other end.

What it does

Our product crawls the store, enriches product data semantically, and generates dynamic JSON optimized for how people actually search (and subsequently, how their agents would behave). The AI Agent predicts likely query keywords and maps products accordingly, making retrieval faster and more accurate. The output is hosted by Mime and accessed via a simple snippet.

How we made it

To build a lean hackathon mvp, we used lovable for the frontend, antigravity and cursor for the backend and firecrawl for web crawling.

Challenges we ran into

Understanding what is the AI Agent shopping process. Defining what the Ai-ready product data should look like

Turning Mime into a plug and play solution

Building a semantic layer that actually improves retrieval

Accomplishments that we're proud of

Building a relatively simple and intuitive solution to a complex problem.

not sleeping throughout the hackathon, and chugging dozen of redbulls :)

What we learned

How AI agents browse and shop, and how GEO affects it

And learning that europe is the hub for the most cracked builders!

What's next for Mime

Becoming the standard infrastructure for agent-driven commerce, through positioning MIME at the center of what’s shaping up to be a multi-billion dollar industry.

Built With

• antigravity
• cursor
• firecrawl
• lovable
• supabase
• typescript