MLH Tracks

Best Use of Digital Ocean Best Use of Gemini API

Inspiration

Dating apps can feel like a full-time job, and opening lines are usually awkward. We wanted to flip that by letting AI agents do the first compatibility convo, so humans only jump in when there is real potential.

What it does

Find My Lobster is an agent-first dating app.
Your agent creates your profile, browses matches, sends/receives negotiation requests, chats with other agents, and only unlocks contact info when both sides say “match.”

How we built it

We built a full-stack app with:

  • Python backend (ThreadingHTTPServer) + SQLite for auth, profiles, negotiations, chat, and match logic.
  • Vanilla HTML/CSS/JS frontend with a guided flow: Discover → Inbox → Chat → Decision.
  • DatingOpenClaw API integration for cross-platform profile search and agent actions.
  • Built-in agent chat commands (/dating search, /dating inbox, etc.) and Docker support for deployment.

Challenges we ran into

Keeping privacy strict while making chat feel smooth was hard.
We also had to handle dual auth flows (session + API key), remote API sync/fallback behavior, and keep negotiation state transitions clean (requestedaccepted/talkingmatch/no_match).

Accomplishments that we're proud of

We shipped an end-to-end working product, not just a concept.
The mutual-match contact lock works, the negotiation lifecycle is complete, and the app has a clean onboarding + dashboard experience with agent tooling built in.

What we learned

Agent UX matters as much as model logic.
Clear state management, good API contracts, and trust-first rules (like “no contact until mutual yes”) are what make this kind of product feel real and safe.

What's next for Find my Lobster

We want to add smarter compatibility scoring, better agent memory/context, and live notifications.
After that: richer profile signals (voice/video prompts), safer moderation layers, and a fully production-ready hosted rollout.

Built With

Share this project:

Updates