AITracker
See at a glance how many tokens you use, how much you spend, and which agent tools work best for you.
AITracker is an open-source, local-first AI workspace.
It automatically tracks tokens, costs, and usage trends across AI tools such as Claude Code, Codex, and Cursor; brings Skills and frequently used configurations into one place; distills Skills from real work and usage history; and builds a personal knowledge base and long-term memory. Every AI session can become a reusable capability for the next one.
Open Source · Local · No Account Required
Note: The diagrams are preserved from the source document. Chinese text embedded inside the image files remains unchanged; all Markdown copy in this document is translated into English.
Why AITracker?
The number of AI tools keeps growing.
Claude Code, Codex, Cursor, Cline, Gemini CLI, OpenCode…
As the number of tools increases, so do the questions:
- How much AI am I actually using?
- Where are my tokens and money going?
- Which tool or model is the best fit for me?
- Where are all my Skills and other configurations scattered?
- Do I have to configure everything again whenever I switch AI tools?
- Can I reuse a method that worked today the next time I need it?
- Can AI remember my projects and long-term working experience?
AITracker brings all of this together in one place.
From understanding your AI usage, to managing AI capabilities, to distilling Skills and building lasting knowledge and memory, AITracker turns AI from a one-off tool into an evolving system of reusable capabilities.
Quick Start
AITracker is a desktop application built with Electron, TanStack Start, React,
and TypeScript. Security scanning uses the published
@estelwalks/agent-threat-scanner npm package, so a fresh clone only needs the
root project dependencies.
Requirements
- Node.js 24 or later
- npm 10 or later
- macOS or Windows for the complete desktop experience
git clone https://github.com/estelwalks/aitracker.git
cd aitracker
npm ci
npm run dev:desktop
To run the browser development server only, use npm run dev.
Install
You can download the installer of the latest release directly. These links stay valid across releases and always resolve to the newest published one: installer names carry no version, so the release tag is the only thing that changes.
- macOS (Apple Silicon): AITracker-arm64.dmg
- macOS (Intel): AITracker-x64.dmg
- Windows (x64): AITracker-Setup-x64.exe
- Windows (ARM64): AITracker-Setup-arm64.exe
Each release lists its own release-metadata.json and checksums.txt, which
record the exact version and SHA-256 of the files in that release. For a
byte-exact build, take the file from the release whose tag you need instead of
from latest.
All releases are listed on the Releases page.
The following commands install the official stable release on macOS or Windows. The package-manager entries become available after the stable release and their corresponding distribution metadata have been published.
npx (macOS and Windows)
With Node.js and npm installed, run:
npx --yes @estelwalks/aitracker@latest
This command downloads and opens the desktop installer. @latest follows the
newest stable release, so it never needs editing; pin a version
(@estelwalks/aitracker@1.0.3) when you need to reproduce an exact build. Beta
builds are published under @beta.
If your npm configuration uses a mirror that does not contain the package, use the official npm registry explicitly:
npx --yes --registry=https://registry.npmjs.org @estelwalks/aitracker@latest
Homebrew (macOS)
Install and upgrade the stable Cask from the project's Tap:
brew tap estelwalks/aitracker
brew install --cask estelwalks/aitracker/aitracker
brew upgrade --cask estelwalks/aitracker/aitracker
WinGet (Windows)
Install and upgrade the stable package from the Microsoft Community Repository:
winget install --id estelwalks.AITracker -e
winget upgrade --id estelwalks.AITracker -e
Build and Test
npm run build:desktop # Web app + Electron main/preload bundles
npm run typecheck # App and Electron integration
npm run lint # App lint rules
npm run test:all # Unit, tooling, database, and scanner tests
npm run check:opensource-hygiene
Platform installers can be produced with npm run dist:mac,
npm run dist:win:x64, or npm run dist:win:arm64. Signing and notarization
credentials are not stored in this repository.
The standalone installer launcher is packaged as aitracker and supports
macOS arm64/x64 and Windows x64/arm64 (including Windows on ARM). Linux is not
supported by the launcher.
See Development Guide for the complete command matrix, generated-file policy, and repository layout.
See Privacy for data-handling details.
🌟 Core Capabilities
| Capability | Description |
|---|---|
| AI Usage Analytics | Automatically aggregates tokens, costs, models, projects, and usage trends across AI tools, so you know exactly how much AI you use and where your money goes |
| AI Tool Analytics | Uses your real usage data to compare the frequency, consumption, and trends of different tools and models, helping you find the AI that suits you best |
| Skills Management | Automatically discovers and centrally manages Skills scattered across different AI tools, reducing repeated searching and maintenance |
| Configuration Management | Centrally manages commonly used AI configurations such as Rules, with cross-tool reuse being added progressively |
| Skills Distillation | Extracts reusable experience from real work, historical sessions, and methods that have already been proven to work, gradually turning it into Skills |
| Knowledge Base | Continuously organizes project materials, methods, experience, and valuable information into a personal knowledge base |
| Long-Term Memory | Continuously preserves project context, usage habits, and important experience, so your next AI session does not have to start from scratch |
| Local First | Keeps core data and configurations on your computer whenever possible, with core capabilities available without an account |
Understand Your AI Usage
AITracker automatically collects usage data from AI coding tools on your computer and brings the scattered data together in one Dashboard.
You can view:
- Token usage
- Input / output / cache tokens
- Daily, weekly, and monthly costs
- Usage across different AI tools
- Usage across different models
- Project consumption
- Usage trends
No more guessing how much AI you use.
See how much you use, how much you spend, and which tool works best—all at a glance.
Manage Your AI Capabilities
What is becoming difficult to manage is not just the AI tools themselves, but also the configurations and capabilities scattered across them.
AITracker automatically discovers these configurations and provides a unified management entry point.
You no longer need to remember which directory contains a particular Skill, or reconfigure everything from scratch whenever you switch tools.
Skills Distillation
A truly valuable Skill does not necessarily need to be written from scratch.
AITracker identifies experience worth reusing from your real work, historical sessions, and methods that have already worked, then gradually distills it into Skills.
Turn “I finally got it working this time” into “I can reuse this directly in the future.”
Reuse Across Tools
A useful Skill should not belong only to Claude Code.
A configuration that works should not force you to start over when you switch to Codex, Cursor, or another tool.
AITracker aims to build a capability layer above different AI tools:
Manage everything in one place, then continue using it across different AI tools.