As cycling grows in popularity, many riders turn to platforms like Strava to log their rides. While Strava's free tier automatically syncs with smartwatches and cycling computers, much of the deeper analysis and training planning required for serious athletic progression remains locked behind a paid subscription. However, the availability of a public Strava API offers developers a great opportunity to build custom, self-hosted analytics tools tailored to their specific needs.
To address this, a developer built PerfRide, a performance management web app powered by the Antigravity framework and the Strava API. The project consists of two distinct components: the core analytics application itself, and an intelligent agent that suggests optimal workouts based on the athlete's current physical condition. This post details the architecture and features of the analytics application.
The application was designed with specific goals in mind. For ride and training management, it enables users to review past rides, visualize power and heart-rate distributions by intensity zones, analyze color-coded performance traces, and track overall fitness progress (including metrics like CTL, ATL, and TSB). For ride and training planning, it generates structured training plans based on upcoming event goals and leverages an agent to recommend the next workout depending on the athlete's fatigue and training load.
PerfRide stands out by integrating several highly advanced and experimental features:
1. Dashboard: Integrates directly with Strava to display recent activities, weekly summaries, and a fitness progress chart (CTL/ATL/TSB), alongside detailed single-ride analysis featuring heart rate zones, power profiles, and elevation overlays.
2. Climb Simulator: This feature utilizes a physics-based simulation (factoring in air resistance, rolling resistance, and drivetrain loss) to predict climbing times based on a rider's power-to-weight ratio and real segment data. Users can query segments via a map or import starred segments directly from Strava.
3. Pace Optimizer: Leveraging research from the 2025 Sports Engineering paper, 'A numerical design methodology for optimal pacing strategy in the individual time trial discipline of cycling,' this optimizer calculates the mathematically ideal pacing strategy based on a course's specific profile.
4. Training Planner: This tool works backward from a target race date to automatically generate a phased, periodized training schedule, optimizing the athlete's progression path.
[AgentUpdate Depth Analysis] The development of PerfRide highlights a significant shift in the AI Agent ecosystem toward highly specialized, domain-specific vertical agents. Rather than relying solely on generalized LLM reasoning, this project elegantly combines the lightweight Antigravity orchestration framework with rigorous physical and physiological models, such as CTL/ATL fatigue tracking and aerodynamic resistance equations. This synthesis of 'hard science calculations' and 'agentic planning' represents the next evolutionary step for personal AI. By bridging existing, data-rich APIs like Strava with specialized agentic workflows, developers can bypass the limitations of generic chatbots to deliver hyper-personalized, actionable insights. In the long run, this paradigm will accelerate the proliferation of micro-agents that seamlessly integrate into niche industries, turning raw sensor data into proactive, intelligent decision-making systems.
