What is claude-meter?
Claude-meter is a local proxy that intercepts Claude Code API calls to Anthropic, capturing raw traffic and normalizing it into structured logs for tracking hidden usage limits like 5h and 7d windows. Developers point their Claude Code client at it via ANTHROPIC_BASE_URL, then run Python analysis tools to generate token breakdowns, utilization summaries, interactive dashboards, and rough budget estimates in dollars. Built in Go for the proxy and Python for analysis, it reveals rate-limit signals, cache hits, and model-specific quotas that Anthropic keeps opaque.
Why is it gaining traction?
Unlike basic token counters, it reverse-engineers real quota behavior from live headers, weighting cache reads cheaply and estimating price-based budgets—key for Claude power users hitting invisible walls. The quickstart proxy plus CLI summaries and one-command dashboards make it dead simple to monitor sessions without changing workflows. At 46 stars, it's niche but hooks devs frustrated by "Claude down" surprises, offering concrete tracking over vague Pro/Max gauges.
Who should use this?
Heavy Claude Code users on Max or Pro plans who burn through quotas during long coding sessions and need precise claude usage meter insights. Ideal for AI engineers optimizing model mixes (Opus vs Sonnet) or tracking cache efficiency across projects. Skip if you're a casual prompter—this shines for daily drivers reverse-engineering limits.
Verdict
Grab it now if you're deep in Claude Code and want a claude limit meter; the alpha tools already deliver actionable summaries and charts from real data. With 1.0% credibility from low stars and research-stage polish, treat it as a local observability hack—not production billing—while it evolves toward crowdsourced quota maps. (187 words)
