smart-mcp-proxy/mcpproxy-go

Methodology

MAU is estimated from public GitHub repository data — no telemetry or tracking is used. The composite formula combines four signals:

• Downloads (50% weight): Release downloads in rolling 90-day window × 25% retention rate × 85% deduplication factor. Based on Fred Wilson's 30/10/10 engagement rule, adjusted conservatively for CLI tools where CI/CD pipelines inflate download counts.
• Community / Iceberg (25% weight): Unique issue authors × 100 + unique PR authors × 500, based on the participation inequality principle (90-9-1 rule). For every contributor who files an issue, ~100 users exist silently.
• Traffic (15% weight): GitHub's 14-day unique views & clones extrapolated to 30 days × 50% real-user engagement factor.
• Stars & Forks (10% weight): 5% of stargazers + 30% of forkers estimated as active users. Weakest signal — stars correlate loosely with usage.

A floor is set at 6-month total downloads × 15% retention (install base of “set and forget” tools). The final estimate is max(composite, install_base).

Accuracy: This is an order-of-magnitude approximation. The true value may differ by 2–4× in either direction. Key limitations include CI/CD download inflation, no package manager tracking (Homebrew, apt, etc.), and non-independent signals (a downloader may also star the repo). The confidence interval [0.5×, 2.0×] is a heuristic range, not a statistically derived bound. This estimate is most useful for tracking adoption trends over time rather than as a precise user count.