Running multiple AI agents like OpenClaw? Your API costs can spiral fast, especially when prompt caching makes traditional usage tracking misleading.
StackMeter gives you session-level cost visibility for OpenAI and Anthropic APIs, showing exactly what each agent run costs, which sessions benefit from cached prompts, and how costs break down across machines.
🤖 **AI API Usage Tracking (OpenClaw native)**
• Per-session breakdowns with start/end times
• Cache read/write tracking (see which sessions paid for cache writes vs benefited from reads)
• Multi-machine support (track dev laptop vs production)
• Model-level analytics (compare GPT-4o vs Claude Sonnet costs)
• Effective cost per 1k tokens (accounting for cache savings)
• Privacy-first: monitor usage without sharing API keys
One command to get started:
`npx @stackmeter/cli connect openclaw --auto`
💰 **Full SaaS Stack Tracking**
• Auto-detect tools from GitHub repos (Stripe, Supabase, Vercel, etc.)
• Track fixed subscriptions + usage-based costs in one dashboard
• Budget alerts when spending hits thresholds
• Annual cost projections
📊 **Built for Teams Running Multiple Agents**
When you're running Claude Code, OpenClaw, and background automations, attribution gets messy. StackMeter handles:
- Shared context costs across sessions
- Cross-machine usage aggregation
- Provider comparison (OpenAI vs Anthropic)
- Cost spike detection
**Free tier:** 7-day data retention, unlimited tracking
**Pro ($9/mo):** 90-day retention, data export
Comments (3)
Congrats on the launch
🙌
@blackdwarftech Thanks for the support!
This tool has helped me a lot, so I figured I should publish it for others. Basic usage is free to use!