I Built a Local Analytics Dashboard for Claude Code — Here's What I Learned About My AI Coding Habits

You know that feeling when you check your Anthropic bill and think... where did all those tokens go? I've been using Claude Code daily for months. It's incredible — but it's also a black box. I had no idea which sessions were burning through Opus tokens on trivial tasks, which prompts were triggering tool loops, or how much money I was wasting by not leveraging prompt caching. So I built Claude Code Dashboard — a local-first analytics platform that reads your Claude Code session logs and turns them into actionable insights. Zero telemetry. Zero outbound connections. Your data never leaves localhost. npx claude-code-insights That's it. One command. Dashboard opens at localhost:3838 . The Problem Claude Code stores rich session logs in ~/.claude/projects/ — every message, tool call, token count, and model used. But there's no built-in way to: See how much you're spending per day, per model, per session Detect when Claude gets stuck in a tool loop (calling Bash 12 times in a row) Know if