Honest answers, briefly.

A free, open-source desktop app that turns your Claude Code transcripts into local usage analytics: spend and tokens per model, waste diagnosis, plugin impact, and subscription limit meters. It focuses on why tokens were spent, not just how many.

Nowhere. Artificiency reads the transcript files already on your machine and stores its analysis in a local SQLite database. There is no account, no sign-in, and no upload. Local-first is the design, not a setting.

Yes. The entire codebase is MIT-licensed and public on GitHub: a Rust core, a Tauri v2 shell, and a Svelte 5 dashboard. You can read it, build it yourself, and keep using it free of charge.

Claude Code today, including per-model cost breakdowns and cache accounting. Codex, Gemini, and OpenRouter support is planned. Provider is a dimension, not a fork, so multiple sources will combine into one dashboard.

The app is developed on Linux, and preview builds are published for Linux, macOS, and Windows. The stack is cross-platform by design, and first-class support for all three desktop platforms is the goal.

Preview builds are not code-signed yet, so Windows SmartScreen and macOS Gatekeeper flag the missing signature. Signed, stable release builds per platform are planned. Until then the download page describes the safe way past each prompt.

Totals tell you that you spent tokens. Artificiency also tells you where they leaked: files re-read in the same session, oversized tool outputs, cache churn, and subagent overhead, plus whether a config change you made actually improved efficiency per turn.

Not yet. Artificiency is pre-1.0 and under active construction, so interfaces, the schema, and the UI change frequently. Preview builds are cut automatically with no quality gate. Since everything runs locally, trying it risks nothing but your patience.

Issues and ideas are welcome on GitHub today. Pull requests are not being accepted yet because the architecture is moving too quickly, but that will change as the project stabilises.