Launched this week
agent-usage-guard
Stop Claude Code from spending your usage window in minutes
4 followers
Stop Claude Code from spending your usage window in minutes
4 followers
A Claude Code mod that holds runaway agent spend before it happens: it reads your plan's live 5-hour and weekly usage, then asks before subagent fan-out, heavy-context prompts and retry loops - and refuses when nobody can answer. Local and privacy-minimal (no network, no processes, never stores prompt text), fails open, and /agent-guard report shows what it did. MIT, one plugin install command.
Free
Launch Team / Built With
Framer AI AgentsDesign and publish professional sites with AI
Promoted
Hi Product Hunt, I'm Riccardo, the maker. Thanks @quietlens for hunting it, that was a nice surprise this morning.
I built it after one prompt in June made 394 agent calls. In ten minutes they replayed 84M tokens of context. Over 60 days I hit the session limit 12 times, and it was rarely one big request. It was big contexts sent again and again, with agents multiplying them.
agent-usage-guard is a Claude Code mod. It reads your plan's live 5-hour and weekly usage and asks before:
subagents fan out, from 80% of either window or when too many start at once
a prompt resends 500k+ tokens of context (you can compact first, then send)
a retry loop goes round again after 3 identical failures
If nobody is there to answer, it refuses. It runs locally, makes no network calls and never stores your prompts. If it breaks, it gets out of the way. /agent-guard report shows everything it did.
To install, run these in Claude Code:
What's the most agents you've seen one prompt start?