kiro-usage 0.1.3

Track token usage and costs across Kiro CLI sessions

kiro-usage

Track token usage and costs across Kiro AI coding sessions — in real time.

A lightweight terminal dashboard that monitors your Kiro CLI token consumption and cost estimates — including cache-read/cache-write breakdown — and persists data across /clear and restarts via a background archiver.

Why? Kiro does not expose running token totals in the UI. This tool fills that gap: see exactly how many tokens and dollars each session costs, live, as you work.

---

Features

• Real-time dashboard — live-refreshing terminal UI (powered by Rich)
• Session archiving — background service snapshots sessions every 10 s; data survives /clear and Kiro restarts
• Cache breakdown — separates CacheWrite (new tokens), CacheRead (context resent), and Output tokens
• Cost estimates — cache-aware pricing (5-min write rate vs. 1-hour read rate)
• Multiple time windows — today / last 7 days / last 30 days / all time
• JSON output — pipe-friendly --json flag for scripting
• IDE tracking — tracks usage across Kiro IDE sessions as well as CLI sessions
• macOS + Linux — service registration via launchd (macOS) or systemd (Linux)

---

Install

Requires uv (recommended) or pip with Python ≥ 3.9.

# install uv if you don't have it
curl -LsSf https://astral.sh/uv/install.sh | sh

# install kiro-usage and register the background archiver
uv tool install kiro-usage && kiro-usage install

Or with pip:

pip install kiro-usage
kiro-usage install

---

Usage

kiro-usage              # today's usage, live refresh
kiro-usage week         # last 7 days
kiro-usage month        # last 30 days
kiro-usage all          # all time

kiro-usage --json       # JSON output (machine-readable)
kiro-usage --no-live    # print once and exit (good for scripts/CI)

Service Management

kiro-usage install      # register background archiver (launchd / systemd)
kiro-usage uninstall    # remove background archiver
kiro-usage status       # check archiver status

The background archiver watches the Kiro CLI SQLite database for changes and snapshots sessions to ~/.kiro_sessions/ every 10 seconds. This ensures session data is preserved even after /clear or a Kiro restart.

---

How It Works

Kiro CLI / Kiro IDE
  SQLite database (~/.kiro/...)
        │
        │  file mtime poll (every 10 s)
        ▼
  kiro-usage-archiver  ──►  ~/.kiro_sessions/*.json
  (launchd / systemd)

  kiro-usage viewer    ◄──  ~/.kiro_sessions/*.json
  (on-demand, live-refresh)

Token Metrics

Metric	Description
CacheWrite	New tokens sent this turn (estimated, chars ÷ 4)
CacheRead	Prior context resent to the API (estimated)
Output	Tokens streamed back from the model (from chunks, accurate)
Cost	Cache-aware estimate — write tokens at 5-min rate, read tokens at 1-hour rate

---

FAQ

Q: Does this work with Claude / Anthropic models? Yes — Kiro uses Claude models under the hood. kiro-usage tracks whatever tokens Kiro sends and receives.

Q: Does it need an API key? No. It reads Kiro's local SQLite database directly — no network calls, no extra credentials.

Q: Is data sent anywhere? No. Everything stays local in ~/.kiro_sessions/.

Q: Why do costs look like estimates? Kiro doesn't expose the exact prompt/completion split, so input tokens are approximated from character count (÷ 4). Output tokens come from streaming chunks and are accurate.

Q: Does it work without systemd / launchd? Yes. Run kiro-usage-archiver manually in a terminal, or just use kiro-usage without the background service (you'll only see current-session data without persistence across /clear).

---

Uninstall

kiro-usage uninstall
uv tool uninstall kiro-usage

---

License

MIT