aitrack

Local-first AI token usage tracker for coding assistants (Opencode, Kiro, and more).

Tracks token consumption, estimates costs, and provides live usage monitoring — all on your machine, no cloud needed.

Features

• Discovery — automatically finds AI tool data sources on your machine
• Incremental scanning — avoids duplicate imports with SHA-256 dedup
• Daily/weekly/monthly/lifetime stats — tokens and cost summaries
• Breakdowns — by tool, model, and provider
• Cost calculator — configurable pricing for all major models
• Live dashboard — Textual-based real-time UI
• File watching — auto-rescans when data files change
• Session analytics — top sessions by tokens/duration/cost
• Export — CSV and JSON export for any period

Supported Tools

Installation

# Clone the repo
git clone <url> aitrack
cd aitrack

# Install with pip
pip install -e .

# Or use the Makefile
make install

Requires Python 3.12+.

Quick Start

# Discover AI tool data sources on your machine
aitrack discover

# Import all discovered data
aitrack scan

# View today's usage
aitrack today

# View lifetime stats
aitrack lifetime

# View cost breakdown
aitrack cost

# Launch the live dashboard
aitrack live

# Watch for file changes and auto-rescan
aitrack watch

# Export data for this month
aitrack export csv --period month

CLI Reference

aitrack discover

Detects AI tools and their data sources on the local machine.

Output: tool name, source type, path, parser compatibility, record count.

aitrack scan

Reads all discovered sources and imports usage records. Skips duplicates using content hashing. Creates the SQLite database at ~/.local/share/aitrack/usage.db.

aitrack today / week / month / lifetime

Shows:

• Total input, output, reasoning, cache read/write tokens
• Estimated cost
• Breakdown by tool, model, and provider

aitrack cost

• Pricing configuration for all models
• Cost summary by period (today, week, month, lifetime)
• Cost breakdown by provider

aitrack watch

Monitors Opencode and Kiro data directories for file changes. Automatically rescans when changes are detected (debounced at 5s).

aitrack live

Launches a Textual-based terminal dashboard that auto-refreshes every 5 seconds. Shows:

• Today/week/month/lifetime stats panels
• Model leaderboard
• Tool leaderboard

aitrack sessions

Lists all tracked sessions with duration, token count, request count, and cost.

aitrack top-sessions

aitrack top-sessions --sort-by tokens --limit 10

Sort by: tokens (default), duration, or cost.

aitrack export

aitrack export csv --period lifetime --output ./my_data.csv
aitrack export json --period month

Periods: today, week, month, lifetime.

aitrack status

Quick health check: database path, size, total record count, last activity, and tools tracked.

aitrack reset

Deletes the local usage database. Prompts for confirmation unless --force/-f is passed. Use this to clear bad data and re-scan from scratch.

aitrack reset          # prompts for confirmation
aitrack reset --force  # skips confirmation

Global options

• --version — print the installed version and exit.
• --verbose / -v — enable debug logging to stderr (in addition to the log file at ~/.local/share/aitrack/log/aitrack.log).

Pricing Configuration

Pricing is stored at ~/.config/aitrack/pricing.json. You can edit this file to add or update model pricing.

Default pricing includes:

• GPT-5, Claude Sonnet 4, Claude Opus 4, Claude Haiku 4.5, Gemini 2.5 Pro
• OpenRouter models (catch-all with $0 default)
• Kiro models (auto, claude-haiku-4.5, claude-opus-4.8, glm-5)
• Opencode big-pickle (free)

Database Schema

~/.local/share/aitrack/usage.db

Example Output

$ aitrack today

          Today Usage
╭─────────────┬────────┬───────╮
│ Item        │ Tokens │ Cost  │
├─────────────┼────────┼───────┤
│ Input       │ 70.0K  │ $0.00 │
│ Output      │ 28.7K  │ $0.00 │
│ Reasoning   │ 2.5K   │ -     │
│ Cache Read  │ 2.6M   │ -     │
│ Cache Write │ 0      │ -     │
│ Total       │ 2.7M   │ $0.00 │
╰─────────────┴────────┴───────╯
           Records: 3

             Today by Tool
╭──────────┬────────┬───────┬─────────╮
│ Name     │ Tokens │     % │    Cost │
├──────────┼────────┼───────┼─────────┤
│ opencode │  50.2M │ 99.8% │   $0.00 │
│ kiro     │  84.2K │  0.2% │ $0.3791 │
╰──────────┴────────┴───────┴─────────╯

Architecture

aitrack/
├── src/aitrack/
│   ├── cli.py                  # Typer CLI entry point
│   ├── collectors/             # Data source collectors
│   │   ├── opencode.py         # Opencode DB + log scanner
│   │   ├── kiro.py             # Kiro DB scanner
│   │   └── generic.py          # Generic JSON/NDJSON/log scanner
│   ├── commands/               # CLI command implementations
│   │   ├── discover.py         # Source discovery
│   │   ├── scan.py             # Data import
│   │   ├── stats.py            # today/week/month/lifetime
│   │   ├── cost.py             # Cost summaries
│   │   ├── watch.py            # File watcher
│   │   ├── sessions.py         # Session analytics
│   │   ├── status.py           # DB health check
│   │   ├── reset.py            # Clear local database
│   │   └── export_.py          # CSV/JSON export
│   ├── database/               # SQLAlchemy + Repository
│   │   ├── models.py           # ORM models
│   │   └── repository.py       # Data access layer
│   ├── dashboard/              # Textual live dashboard
│   │   └── live_app.py         # Real-time UI
│   ├── models/                 # Pydantic models
│   │   ├── usage_record.py     # Usage data schema
│   │   ├── discovery.py        # Discovery schema
│   │   └── pricing.py          # Pricing schema
│   └── utils/                  # Utilities
│       ├── pricing.py          # Cost calculator
│       ├── formatters.py       # Rich terminal formatting
│       └── logging.py          # Logging setup
├── tests/                      # Test suite (57 tests)
├── pyproject.toml
├── Makefile
└── README.md

Platform Support

Currently tested on Linux (~/.local/share, ~/.config paths). macOS uses different standard paths (~/Library/Application Support, ~/Library/Preferences) which are not yet auto-detected — contributions welcome.

Testing

make test        # Run all tests
make lint        # Lint with ruff
make format      # Format with ruff

Development

make dev    # Install with dev dependencies
make build  # Build distribution packages
make clean  # Clean build artifacts

See CONTRIBUTING.md for details on adding new collectors and the development workflow.

License

MIT — see LICENSE.