onWatch is a Go CLI that runs as a background daemon, polls your API quotas (Anthropic, OpenAI Codex, GitHub Copilot, Synthetic, Z.ai, Antigravity), stores history in SQLite, and serves a Material Design 3 web dashboard.
Key decisions:
- Single binary (~13MB), no runtime dependencies - <50MB RAM with all 6 providers polling in parallel - All data stays local — zero telemetry, no cloud - One-line install for macOS, Linux, Windows - Docker support (distroless, non-root, ~10MB image)
The insight I kept coming back to: knowing your current usage isn't enough. You need historical cycle data — which sessions burn quota fastest, how usage compares across billing periods, and when to expect resets.
Install: `curl -fsSL https://raw.githubusercontent.com/onllm-dev/onwatch/main/install.sh | bash`
https://github.com/onllm-dev/onwatch