LLM Cost Attribution Tool with Network-Layer Monitoring

Halton Meter meters LLM API spend at the network layer. A local daemon sits at the system proxy (mitmproxy with a locally trusted root CA), captures every request to Anthropic, OpenAI, Google and xAI, prices it from a versioned rates corpus, and attributes cost to project, machine and person. Because capture happens below the SDK, it works with any client, including tools that expose no hooks at all, such as Claude Code on a Max subscription. No competitor meters that surface.