Artificiency

Flarepoint | Artificial Efficiency. Local-first analytics for your AI coding usage.

Website

Most usage trackers tell you how much you spent. Artificiency tells you why. It shows where the tokens actually went (re-read files, oversized tool dumps, cache churn, subagent overhead), and whether the change you made (a new plugin, an edited CLAUDE.md, a different workflow) actually helped. It runs entirely on your machine, and your transcripts never leave it.

⚠️ Early development. Artificiency is pre-1.0 and under active design and construction. Interfaces, the data schema, and the UI change frequently, and there are no stable releases yet. To run it today, grab a preview build or build from source (both below).

---

Screenshots

Overview	Waste diagnosis

The dashboard is under active visual development, so UI elements are subject to change.

---

What it does today

• Overview: spend, tokens, turns, sessions and cache activity over any range (1h → all time), a per-model breakdown with estimated cost, and a trend chart you can flip between metrics (spend, output, turns, cache, and so on).
• Waste diagnosis: redundant file re-reads, largest single tool outputs, and output volume per tool, each with an estimate of the tokens involved and tips on avoiding it.
• Current usage: your subscription limit meters (session and weekly), read from the same source the official view uses.
• Cost engine: per-model pricing (input, output, and cache read/write) computed locally.

Status & roadmap

Artificiency is being built on Linux, which is the daily development target for now. The stack (Tauri v2 with a Rust core) is cross-platform by design, and the goal is full support for all three desktop platforms and many AI providers.

Area	Today	Planned
Platform	Linux	macOS and Windows as first-class targets.
Provider	Claude	Codex, Gemini, OpenRouter and others. Provider is a dimension, not a fork, so multiple sources combine into one dashboard
Distribution	Unsigned preview builds, or build from source	Signed, stable release builds per platform

macOS and Windows are not officially supported yet, but the code is written to run there. If you would like to try it early you can grab a preview build or build it yourself.

---

Download

Preview builds for Linux, macOS, and Windows are published on the Releases page. Download the installer for your platform and run it.

⚠️ Preview builds are unstable and untested. They are cut automatically from a tagged commit, with no quality gate. Expect bugs, breaking changes between versions, and features that appear and disappear. Everything runs locally, so the only thing at risk is your patience, not your transcripts.

The builds are not code-signed, so your system will warn you the first time you open one:

• macOS: right-click the app and choose Open, then confirm. A plain double-click stays blocked on unsigned apps.
• Windows: on the SmartScreen prompt, click More info, then Run anyway.
• Linux: make the AppImage executable (chmod +x), or install the .deb or .rpm.

Would you rather build it yourself? See Build from source below.

---

Build from source

You will need, on every platform:

• Rust (stable, via rustup)
• Node.js 18+ (CI uses 22)

Plus the platform-specific system dependencies for a Tauri v2 app:

Linux

The WebKit webview and supporting libraries.

# Arch / CachyOS
sudo pacman -S --needed webkit2gtk-4.1 base-devel curl wget file openssl \
  librsvg libayatana-appindicator gtk3

# Debian / Ubuntu
sudo apt install libwebkit2gtk-4.1-dev build-essential curl wget file \
  libxdo-dev libssl-dev libayatana-appindicator3-dev librsvg2-dev

On GNOME, the tray icon needs a shell extension. The app is dashboard-first and does not depend on the tray.

macOS

xcode-select --install   # Command Line Tools (provides the required toolchain)

The WebView is provided by the system (WKWebView), so there is nothing else to install.

Windows

• Microsoft C++ Build Tools (the "Desktop development with C++" workload).
• WebView2 runtime. Preinstalled on Windows 11 and current Windows 10. Otherwise, install the Evergreen runtime from Microsoft.

Then:

npm install

# Run the desktop app in development (hot-reloads the UI)
npm run tauri dev

# Produce a release build + installer for your platform
#   output lands under src-tauri/target/release/bundle/
npm run tauri build

First run backfills your existing Claude Code transcript history from ~/.claude/projects, so the dashboard is populated immediately.

---

Project layout

• crates/artificiency-core is the Rust core: the SQLite store, collectors, and the cost/stats engine. It has no GUI dependencies, and compiles and tests standalone.
• src-tauri is the Tauri v2 desktop shell exposing core commands to the UI.
• src is the Svelte 5 dashboard.

Development

cargo test -p artificiency-core            # core unit tests
npm run check                              # type-check the frontend
cargo run --example backfill /tmp/smoke.db # headless collector smoke run
cargo run --example limits                 # print what the usage widget would show

CI runs the core test suite on Linux, macOS, and Windows, and type-checks and builds the frontend on Linux.

---

Contributing

Thanks for the interest. Artificiency is still in its design and early-construction phase, and the architecture, schema, and UI are moving too quickly for outside contributions to be practical right now, so pull requests are not being accepted yet. That will change as the project stabilises. In the meantime, issues and ideas are welcome.

License

MIT © 2026 Floran Prins

---

Artificiency is a Flarepoint project. Local-first, and staying that way. Your data never leaves your machine.