Developer Vibe-Codes a macOS AI Usage Dashboard

A developer frustrated by constantly checking multiple AI service dashboards has built UsageBoard, a sleek macOS menu bar application that consolidates API usage tracking for services like OpenAI Codex, DeepSeek, and MiniMax into a single interface. The project, shared on a popular Chinese developer forum, exemplifies the growing 'vibe coding' movement — where developers leverage AI assistants to rapidly prototype and ship functional software.

The creator, who subscribes to multiple AI services for daily development work, described the pain of visiting numerous websites multiple times per day just to monitor token consumption and billing. Rather than continuing to context-switch between dashboards, they decided to build a purpose-built native macOS app to solve the problem once and for all.

Key Takeaways

• UsageBoard is a free macOS menu bar app that aggregates AI API usage data from multiple providers
• Currently supports 5 services: OpenAI Codex, ZhiPu AI (GLM), DeepSeek, MiniMax, and Tavily
• Features a plugin architecture allowing independent configuration per service
• Built entirely through 'vibe coding' — AI-assisted rapid development
• Includes token usage charts, auto-refresh, disk caching, and drag-and-drop plugin ordering
• Supports both percentage-based and numeric usage displays with color-coded progress bars

The Problem Every AI Developer Faces

API usage tracking has become a genuine productivity drain for developers working across multiple AI platforms. Unlike the early days of cloud computing where most developers worked within a single ecosystem like AWS or Google Cloud, today's AI landscape demands multi-provider strategies.

A typical AI-savvy developer in 2025 might use OpenAI's Codex for code generation, DeepSeek for cost-effective reasoning tasks, and specialized services like Tavily for AI-powered web search. Each of these services has its own dashboard, its own billing cycle, and its own way of displaying usage metrics.

The cognitive overhead is real. Developers report spending 15 to 30 minutes daily just checking various dashboards to avoid surprise bills or hitting rate limits mid-project. This is the exact friction that UsageBoard aims to eliminate — a single glance at a menu bar icon replaces half a dozen browser tabs.

Inside UsageBoard's Architecture

What makes UsageBoard particularly interesting from a technical standpoint is its plugin-based architecture. Rather than hardcoding support for each AI service, the developer built a modular system where each provider is implemented as an independent plugin.

Each plugin can be configured with its own refresh interval and parameters. The settings interface is automatically generated from script metadata, meaning adding support for a new AI service requires minimal boilerplate code. Key architectural features include:

• Menu bar residency with a quick-preview popup on icon click
• Grouped and tabbed display modes for organizing multiple services
• Manual, timed, and per-card refresh options for granular control
• Disk-based caching by state ID, so the app displays last-known data on launch
• Remote icon loading with local caching for plugin branding
• Auto-installation of built-in plugins to the user plugin directory on first launch

The app currently provides detailed token usage charts for 2 services — Codex and ZhiPu AI. For ZhiPu, the charts pull data directly from the official API. For Codex, the developer took a different approach, analyzing local conversation sessions to reconstruct usage patterns. This hybrid strategy demonstrates pragmatic engineering — using whatever data source is most reliable for each provider.

What Is Vibe Coding, and Why Does It Matter?

Vibe coding is a term coined by Andrej Karpathy, former Tesla AI director and OpenAI researcher, to describe a development style where programmers use AI assistants to write most of the code while they focus on high-level direction and 'vibes.' Rather than meticulously crafting every function, vibe coders describe what they want, review AI-generated output, and iterate rapidly.

This UsageBoard project is a textbook example of vibe coding done right. The developer identified a clear personal pain point, described the desired functionality, and used AI coding tools to produce a polished, feature-rich macOS application. The result is not a hacky prototype — it includes sophisticated features like drag-and-drop plugin ordering, subscription tier badges with styled labels, startup launch configuration, and an integrated update system.

The vibe coding approach is reshaping how individual developers think about project scope. Features that would have taken a solo developer weeks to implement — like auto-generated settings forms, disk caching layers, and plugin architectures — can now be scaffolded in hours with AI assistance. This dramatically lowers the barrier to building professional-quality tools for niche problems.

How UsageBoard Compares to Existing Solutions

Several tools already exist in the API management space, but most target enterprise users or focus on a single provider. Helicone, for instance, provides detailed OpenAI usage analytics but requires routing API calls through a proxy. LangSmith by LangChain offers comprehensive LLM observability but is designed for production monitoring rather than personal subscription tracking.

UsageBoard occupies a different niche entirely. It is:

• Lightweight — a menu bar app, not a web platform
• Multi-provider — supports 5 services out of the box with room to grow
• Privacy-conscious — queries official APIs directly rather than proxying traffic
• Developer-first — designed for individual developers, not teams or enterprises
• Extensible — new plugins can be added without modifying the core application

Compared to simply bookmarking 5 dashboard URLs, UsageBoard offers real-time data aggregation, visual progress indicators, and automatic refresh — turning a manual chore into passive awareness. The subscription tier badge feature is a particularly nice touch, instantly reminding users which plan they are on for each service.

The Broader Trend: Developer Tooling Goes Personal

UsageBoard reflects a larger shift in the developer tools ecosystem. As AI services proliferate — there are now over 100 commercially available LLM APIs — developers increasingly need personal infrastructure to manage their AI stack.

We are seeing this trend manifest in several ways. Developers are building custom CLI tools to switch between AI providers based on task complexity. Teams are creating internal dashboards to track spending across OpenAI, Anthropic, Google, and open-source model hosting. And now, individual developers are crafting native desktop apps to monitor their personal AI consumption.

This is reminiscent of the early 2010s when developers built personal dashboards for GitHub contributions, server uptime, and social media metrics. The difference is that today's tools can be built in a fraction of the time thanks to AI-assisted development. A virtuous cycle emerges: AI tools help developers build better tools for managing their AI tools.

What This Means for Developers and the AI Ecosystem

For individual developers, UsageBoard solves an immediate and growing problem. As AI API costs continue to vary widely — DeepSeek offers tokens at a fraction of OpenAI's pricing, while specialized services like Tavily charge premium rates for search capabilities — cost awareness becomes critical to sustainable AI-powered development.

For AI service providers, projects like UsageBoard signal a demand for better, standardized usage APIs. The fact that the developer had to use local session analysis for Codex usage data, rather than a clean API endpoint, highlights gaps in provider tooling. Companies that make their usage data easily accessible via well-documented APIs will have an advantage in developer experience.

For the broader ecosystem, this project demonstrates that vibe coding is maturing beyond toy projects and demo apps. A menu bar application with plugin architecture, disk caching, auto-updates, and multi-provider support is a legitimate piece of software — and it was built by a single developer using AI-assisted workflows.

Looking Ahead: The Future of AI Usage Management

UsageBoard currently supports 5 AI services, but the plugin architecture makes expansion straightforward. The most obvious candidates for future integration include Anthropic (Claude API), Google (Gemini API), Mistral AI, and Cohere. As the developer has open-sourced the plugin system, community contributions could rapidly expand provider coverage.

The concept could also evolve beyond simple usage tracking. Future versions might include cost forecasting based on usage trends, automated alerts when approaching billing thresholds, or even intelligent routing suggestions — recommending cheaper providers for specific task types based on historical usage patterns.

As AI subscriptions become as common as SaaS subscriptions were a decade ago, tools like UsageBoard may evolve from nice-to-have utilities into essential developer infrastructure. The developer who built this app was scratching their own itch, but they may have stumbled onto a product category that millions of AI developers will eventually need.

For now, UsageBoard stands as a compelling example of what happens when developer frustration meets AI-powered productivity: a polished, practical tool built at the speed of thought.