TextFly | Devpost

Inspiration

We were inspired by recent events involving widespread flight delays and cancellations caused by the ongoing government shutdown. To address this issue, we set out to create an AI-driven solution that operates entirely through text messaging—no apps, no downloads, just seamless communication through your phone’s native messenger.

What it does

With TextFly, you can simply message textfly@tech and get instant, accurate responses to your flight- or travel-related questions — all through text, with no app or downloads required. TextFly goes beyond standard LLMs by using our proprietary multi-agent pipeline, which allows different AI agents to interact with specialized tools. This enables it to: Predict flight delays using real-time Twitter data, Check weather conditions across cities to help you avoid risky travel plans and Provide comprehensive travel insights, all through a simple text conversation

How we built it

We built TextFly using the Photon SDK, which enabled us to seamlessly handle incoming and outgoing user messages. At the core of our system is an orchestrator agent, the main controller responsible for managing the entire conversation flow. The orchestrator agent coordinates with several specialized sub-agents, each equipped with access to various MCP tools. These sub-agents can be called in parallel to handle specific tasks, such as data retrieval or analysis. When appropriate, the orchestrator can also choose to handle a request independently without invoking any sub-agents, ensuring efficient and adaptive performance.

Challenges we ran into

We had a challenge making MCP's on Dedalus. The ones we made only functioned locally not online. Even after attempting to publish the MCP's we made, they couldn't be published.

Accomplishments that we're proud of

The team is proud of building the MCP integration that powers the project’s backend connections. They’re also proud of designing and developing a responsive landing page that clearly communicates the product’s vision. As a team, they took initiative in leading collaboration across multiple components and ensuring all parts came together smoothly. Most of all, they’re proud of successfully completing the project from start to finish.

What we learned

We learned how to create AI agents. We learned how to make MCP's. We learned how to make diagrams for complex agentic workflows. Specifically ones that can interact in almost real time applications, as well as deal with unpredictable inputs.

What's next for Textfly

To explore monetization routes using targeted ads. Think how google charges companies to be placed higher in search. We want to expand over to hotel bookings and other travel bookings.