shanggqm/codexU

What it solves

codexU is a macOS utility that solves the problem of fragmented and opaque AI usage tracking for developers using OpenAI Codex and Claude Code. It eliminates the need to constantly switch between browsers or terminals to check remaining quotas, token consumption, and task progress, providing a centralized, local-first dashboard for monitoring AI workforce productivity.

How it works

The application runs as a macOS menu bar and desktop app that reads data locally from the user's machine. It accesses the Codex app-server for account quotas and reads local SQLite databases and JSONL logs (from ~/.codex/ and ~/.claude/) to calculate token usage, track active tasks, and analyze model performance. All processing is done on-device to ensure privacy, with no usage data uploaded to third-party services.

Who it’s for

• Developers who frequently use OpenAI Codex, Codex CLI, or the Codex desktop app.
• Users who simultaneously employ both Codex and Claude Code and want a single entry point for usage statistics.
• ChatGPT Pro/Team users who need to monitor 5-hour and 7-day quota windows and reset times.

Highlights

• AI Leadership Scoring: A unique metric that evaluates the scale of the AI workforce being managed based on agent lifecycles, parent-child relationships, and autonomous operation evidence.
• Local Inference Monitoring: Tracks real-world performance (P50/P90 latency and effective throughput) aggregated by model and reasoning intensity.
• API Value Estimation: Calculates the "Wool Progress" (API equivalent value) by converting local token usage into USD based on official OpenAI API pricing.
• Unified Dashboard: Provides a comprehensive view of token heatmaps, project rankings, tool/skill usage TOP lists, and a task board for both Codex and Claude Code.
• Local-First Privacy: Operates entirely on the Mac, ensuring that logs, paths, and account data never leave the device.