Overview

Thrugo is an AI-powered platform - available on both mobile and web - that transforms the way people discover and plan events. Instead of scrolling through endless lists, users simply start a conversation. The AI listens, asks the right questions, and dynamically suggests events that match preferences for time, location, and genre. Once an event is selected, users can buy tickets, explore similar options, or even open a group chat where friends can coordinate, vote, and plan together.

Airnauts turned Thrugo from concept into a production-ready MVP, blending user-friendly design with advanced AI architecture to create a platform that feels both intuitive and powerful.

Challenges in a Fast-Moving AI Landscape

Thrugo was built during one of the fastest-moving periods in AI history. At every stage of development, new tools, frameworks, and techniques were emerging. We stayed ahead of the curve, adapting our architecture as capabilities evolved — a process that required both technical creativity and constant iteration.

One of the biggest hurdles was dealing with dynamic event data. Event listings are constantly changing and spread across multiple providers. To ensure the AI could recommend the most relevant, up-to-date experiences, we had to design systems that pulled, filtered, and aggregated data in real time.

Another challenge was striking the right balance between natural conversation and structured data collection. The AI needed to gather essential parameters like date, time, location, and genre without making the experience feel like a rigid questionnaire. Achieving this required careful prompt and context engineering, so that conversations felt fluid while still producing structured, actionable results.

At the same time, the project began when LLM function calling was a brand-new, barely documented feature. MCP servers and modern tool ecosystems hadn't yet arrived. To move forward, our team engineered custom solutions that allowed the AI to retrieve and filter data in the background during conversation. Later, as the field matured, those early workarounds naturally aligned with today's agentic architectures and tool-based approaches.

Ensuring that the AI behaved reliably was another ongoing task. We introduced guardrails to keep conversations safe and goal-aligned, while also managing agent memory so that the AI could sustain context over multiple turns. The entire system was built on a real-time, event-driven architecture using frameworks like LangChain, LangGraph, and Pydantic, ensuring the platform was both flexible and scalable.

Airnauts' Approach

From the beginning, our goal was to create an AI that was autonomous, context-aware, and responsive - one that could handle multi-turn conversations, pull data dynamically, and adapt to evolving user needs.

We invested heavily in conversational design, making sure the AI could ask the right questions at the right moments without overwhelming users. By fine-tuning prompts and carefully engineering context windows, we ensured that Thrugo felt like a helpful guide rather than a scripted survey.

Our technical approach emphasized modularity and adaptability. Even as we worked around early limitations in function calling, we structured the system so it could easily adopt new methods as the ecosystem evolved. This forward-looking design meant that when tools, agent frameworks, and MCP servers became widely available, Thrugo's architecture was already prepared to integrate them seamlessly.

Outcome

Thrugo launched as a production MVP across mobile and web, showcasing how conversational AI can reshape event discovery and planning. Users can now find events in a natural, human-like way, then seamlessly transition into group planning with their friends. With features like pinned events, group chats, attendance voting, and reminders, Thrugo makes coordination effortless.

For Airnauts, Thrugo is proof that with the right approach, conversational AI can move far beyond novelty - delivering real-world utility while maintaining a natural, engaging user experience.

Lessons Learned

Building Thrugo highlighted some of the most important truths about working with AI today. First, that*flexibility is everything: architectures must be designed to adapt as the underlying technology evolves. Second, that **conversation is both an art and a science**: natural dialogue and structured data collection are not opposing goals, but must coexist in careful balance. And third, that early innovation pays off: by working through the limitations of immature tooling, we positioned Thrugo to align naturally with the best practices that have since become industry standards.

Closing Thought

For Airnauts, Thrugo was more than just another project. It was an opportunity to push the frontier of applied AI, experimenting with new techniques while keeping sight of real user needs. From engineering early function-calling workarounds to orchestrating agent autonomy with safety, Thrugo stands as an example of how our team approaches AI: bold, adaptive, and deeply technical - but always in service of creating meaningful products.

‍