Show HN: agent-usage-guard – a Claude Code mod to protect your usage window

Category: devtools

Tags: claude-code, cost-control, ai-agents, guardrails, developer-tooling

Score: 7.0/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 6)

agent-usage-guard is a Claude Code mod that prevents runaway token spend by intercepting agent spawns, heavy-context prompts, and retry loops before they execute. It reads live plan usage, enforces cross-session budgets, and asks the user (or refuses) when limits are hit, with detailed reporting via /agent-guard report. It addresses a real, documented failure mode of multi-agent coding assistants and ships with unusually thorough documentation and a demo.

Target audience: AI-assisted developers, Claude Code users, platform/DevOps engineers managing LLM spend

Repository: https://github.com/rbartoli/agent-usage-guard · TypeScript · MIT · 3 stars

View on Hacker News