Twogether

Your AI planner assistant for date night. Partner agents learn from the couple and the planning agent builds the perfect plan you'll both love.

Comment

Inspiration

Every couple knows the standoff: "Where do you want to eat?" "I don't know, where do you want to eat?" We built Twogether to eliminate that negotiation entirely. let AI handle the compromise so both partners just have to show up to an amazing date.

What it does

Each partner texts their own AI agent, gets privately asked about their mood, budget, and preferences, and a third orchestrator agent synthesizes both responses into a complete date plan which is delivered back to both phones over SMS.

How we built it

Three-agent architecture built on Gemini with an MCP Mac messaging tool. Partner agents conduct adaptive multi-turn "interviews" over text. The orchestrator merges both preference profiles, queries Google Maps for venues, checks weather, and generates a ranked date plan. The entire UX is the messaging thread with a pretty dashboard for the demo.

Challenges we ran into

Coordinating two asynchronous SMS conversations and knowing when both partners were "done" so the orchestrator could fire was trickier than expected. Getting Gemini to reliably output structured data from the orchestrator for downstream API calls also took iteration. We also ran into issues with Twilio - e weren't able to get our number provisioned in time, so we opted for a Mac messaging MCP tool instead.

Accomplishments that we're proud of

Building an end-to-end multi-agent flow with real SMS in and real date plan out. This all with no app required. The private questionnaire mechanic felt like a real product insight.

What we learned

Multi-agent systems with async human inputs are as much a timing and coordination problem as an AI problem. Also: the best AI UX is sometimes no UI at all. In addition we learned many time management skills as this hackathon was an interesting experience to say the least.

What's next for Twogether

Twillio integration, live Resy booking, Beli integration for taste-profile-aware recommendations, a shared date log that feeds the agents' memory over time, and a travel mode that auto-spins up city-specific plans when a couple lands somewhere new.

Built With

Share this project:

Updates