Llmkit

If you're running Claude Code, Cline, or Cursor and want to know what your AI sessions actually cost, this server gives you 11 tools to track spend across providers. Six tools query the LLMKit proxy for budget status, session summaries, and cost breakdowns by key or time range. Five local tools parse Claude Code and Cline state files directly to show per-project costs, cache hit rates, and spending forecasts without needing an API key. The SessionEnd hook logs a cost summary when your coding session ends. Useful when you're burning through tokens on complex refactors and want real numbers instead of guessing, or when you need to justify API spend to your team with per-session breakdowns.