AgenTinder

Home screen, Sage agent selected
After searching for sauna sessions
Matched with HaloHeat

Inspiration

AI Agents are growing in popularity and only becoming more specialized, which means a single agent cannot have every tool, capability, or piece of information it might need. I wanted to explore what happens when an agent receives a request it cannot complete on its own.

What it does

AgentTinder allows users to choose between specialized agents, including a Wellness Agent and a Catering Agent, and give them a request. The selected agent determines what outside capabilities it needs and searches GoDaddy’s ANS for compatible A2A agents. The discovered agents are displayed as Tinder-style cards showing their capabilities and information. When a match is made, AgentTinder connects to the selected agent and uses its response to complete the user’s original request.

How we built it

Whole backend is Flask. Frontend is HTML/CSS/Vanilla Javascript The core workflow is:

The user selects a starting agent and submits a prompt.
The application identifies the capabilities needed to complete the request.
GoDaddy ANS is searched for agents offering those capabilities.
Relevant agents are transformed into interactive AgentTinder cards.
The user chooses to pass or match.
After matching, the selected agent is contacted and its response is used to fulfill the original request.

Challenges we ran into

The biggest challenge I faced was finding potential A2A candidates. Unlike a traditional API, there is no known endpoint, AgenTinder needs to dynamically discover services. Another challenge I faced was honestly keeping the project small. For this being my first hackathon, I needed to take my medicine and keep the project small. AgenTinder could have gone many directions.

Accomplishments that we're proud of

One accomplishment that I am proud of is getting the A2A connected finally and seeing the final answer to the prompt go through. It was tricky.

What we learned

AgenTinder taught me a ton about Agent to Agent communication. Before the hackathon I had no Idea what to build, I attended the GoDaddy workshop and got interested immediately because it is something I have never worked with. I'm glad I took the chance because I learned a ton on how ANS works and the importance of it.

What's next for AgenTinder

What's next for AgenTinder is to expand the base agents and improve scope of ANS search. In a small hackathon project, it was hard getting more than 5 different potential services to communicate with. In the future, I could expand prompt scope to get better results.