Inspiration

Our inspiration comes from the common struggle with time management fatigue and the limitations of traditional to-do list apps. Existing tools force users to manually break down tasks, estimate durations, and constantly reshuffle their calendar when things change. We were inspired to build a truly agentic scheduling assistant that leverages the power of Large Language Models (LLMs) to automate the cognitive load of daily planning and real-time adaptation. We envisioned an app that acts as a proactive, personalized chief of staff, not just a static calendar.

What it does

Promptly is an AI-powered smart daily planner that uses Natural Language Input and a sophisticated LLM to automatically create, manage, and dynamically re-optimize your daily schedule in Google Calendar.

It takes high-level goals (e.g., "Write a paper, hit the gym, and grocery shop") and transforms them into time-blocked events on your calendar, intelligently allocating time based on:

Calendar Availability: Finding the best open slots. Task Properties: Duration, priority, and deadlines. Multi-day Planning: Split up large tasks into smaller and specific daily blocks. Crucially, it acts as a Live Rescheduling Agent, instantly adapting and updating your Google Calendar if a meeting runs long or you fall behind schedule.

How we built it

We built Promptly using a modern, scalable, and agent-focused stack:

Core Logic: The "brain" is an agent running on GPT 4.1, which handles the natural language parsing, intelligent time allocation, priority-based scheduling, and dynamic re-optimization logic. Backend: A Python FastAPI service manages the agent requests, business logic and user data. A REST API hosted on Vercel converts user inputs into tasks that are time-alloted and formatted into calendar-friendly JSON objects. Database: We leveraged Supabase for our backend database and its authentication features to manage user accounts and task persistence. Additionally, we used Firebase to store user data metrics including user preferences and task completion rates. Calendar Integration: We relied heavily on the Google Calendar API to read availability, create new time-blocked events, and execute real-time updates and deletions during rescheduling. Frontend: The user interface was developed with React Native and Expo for a seamless but feature-packed mobile experience with an easy to learn and modern UI. The app features a clean tab-based navigation with dedicated screens for planning, calendar management, and settings. Authentication: We implemented Google Sign-In integration to securely authenticate users and manage their calendar access permissions. Audio Processing Integration: Implementing Whisper speech-to-text required careful handling of audio file formats, real-time transcription processing, and seamless integration with our existing natural language processing pipeline.

Challenges we ran into

LLM Tool Reliability: Ensuring the LLM agent reliably and consistently formatted API calls (specifically the JSON structure for Google Calendar events and updates) was a significant challenge that required meticulous system prompting and validation layers in the FastAPI service. Google Authentication and Scope Management: Integrating with the Google Calendar API required careful management of the OAuth 2.0 flow, specifically requesting and handling the necessary scopes for creating, updating, and deleting events without over-requesting permissions, which was a complex and time-consuming process to implement and test securely. Expo Development Workflow: Managing the transition from Expo managed workflow to development builds required careful configuration of native dependencies and proper setup of the iOS development environment.

Accomplishments that we're proud of

Seamless Natural Language Planning: We achieved a truly friction-free planning experience with our "Planner Chat" feature, where users can simply speak/type a few sentences and have a complete, intelligent, and time-blocked schedule generated in seconds. The Rescheduling Agent: Successfully building and deploying the real-time, adaptive rescheduling logic that automatically updates the user's Google Calendar without manual intervention. This is a core differentiator that solves a major pain point of daily life. Voice-First Planning Experience: Successfully integrated Whisper speech-to-text to create a hands-free planning experience, allowing users to dictate their daily goals and have them automatically converted into structured calendar events. Comprehensive Testing Framework: Built a robust testing system with multiple notification and scheduling components to ensure reliability across different scenarios and app states. What we learned We learned that the true power of LLMs in a productivity tool is not just in generating text, but in complex, multi-step reasoning and automated execution via tool-calling (in our case, the Google Calendar API). This "Agentic AI" approach allows us to move beyond simple automation to dynamic, contextual, and proactive assistance. We also learned the paramount importance of user data privacy and the need for clear consent when integrating with sensitive services like a user's calendar. Additionally, we discovered the importance of having comprehensive testing frameworks in place when building complex scheduling systems.

What's next for Promptly

Deep Integration with Other Tools: Integrate with popular task management platforms (Todoist, Notion, Asana) and communication tools (Slack, Teams) to automatically pull in tasks and set contextual statuses. Team Scheduling & Collaboration: Introduce a feature for teams to schedule collaborative work or meetings, where Promptly can find optimal times that respect all team members' focus time and individual energy curves. Advanced Notification System: Implement a comprehensive notification system that provides intelligent reminders and follow-ups to keep users on track with their scheduled tasks.

Built With

• expo.io
• javascript
• openai
• python
• react-native
• supabase
• typescript
• vercel