"I used to burn through my weekly cap by Wednesday and spend the rest of the week locked out. The pace ring shows me where I'll land by Sunday if I keep this rate, so I throttle myself before Anthropic does it for me."
The Usage Meter for AI Agents.
Track Claude and Codex quota, pacing, cache health, and activity in real time. Built for developers and their agents.
Built for developers who ship with AI
Everything on one screen
Quota pacing, session health, model breakdown, activity liveness, hourly patterns — all updating in real time. Drag, resize, and pin widgets to build your perfect control panel.
Know exactly where your quota stands
Three concentric rings show pace projection, weekly usage, and session usage at a glance. Budget delta tells you how much runway you have today.
Optimize every token
Track cache hit ratios per session. Spot cold starts instantly. Know exactly where tokens are being wasted and where you're saving.
Know when you code hardest
Three time horizons show your peak hours at a glance. Dock-style badges magnify on hover. Bars reveal per-hour message counts across all-time, this week, and today.
Built for agents, not just humans
Every widget has a matching CLI endpoint. Your agents can query quota, activity, and cache health programmatically via JSON over localhost.
"provider": "claude",
"projects": [
{
"name": "data-pipeline",
"seconds_ago": 12,
"state": "active",
"activity_type": "tool_call",
"tool_calls": 3
}
],
"seconds_since_any_activity": 12,
"any_stalled": false
}
$ emusage analytics # today, 14-day trend, all-time
$ emusage cache # per-session cache hit ratios
$ emusage activity # per-project liveness
$ emusage activity --provider codex
See why developers keep EMUsage open all week
The strongest reactions so far are about pacing clarity, catching weekly-cap burn before it locks them out, and giving agents a safe budget check before longer runs.
"My 5-hour windows started draining in 90 minutes and I had no idea what was eating them. Seeing the live burn rate means I catch a hot session in the first 10 minutes now, not when I'm staring at a lockout screen."
"The reset clock in the terminal is buried and the times shift around. Having the countdown sitting on my second monitor means I actually plan deep work around it instead of guessing when my week resets."
"Before this I'd kick off a long agent run and find out two hours later I'd cooked the rest of the week. The localhost endpoint lets us gate longer runs against remaining budget — that alone paid for it."
"The pace ring is the feature for me. It tells me where I should be by tonight and where I'll actually land at this rate. Once you've seen that, working without it feels like coding with the brightness off."
"Cache health surfaced a prompt pattern I'd been running for weeks that was getting almost no reuse. Fixing it cut my daily burn by about a third — I'd never have spotted it from the CLI."