Claude Usage Tracker

Product Introduction

1. Definition: Claude Usage Tracker is an open-source, local-first analytics dashboard and native macOS utility designed to aggregate and visualize Claude AI token consumption and financial expenditure. Technically classified as a cross-tool usage aggregator and cost-monitoring agent, it functions by scanning local session metadata and log files generated by various AI development environments. Unlike traditional cloud-based SaaS tools, it operates entirely on the user's machine using Node.js for data collection and a native Swift-based WKWebView wrapper for the interface.

2. Core Value Proposition: The primary purpose of Claude Usage Tracker is to eliminate "usage blindness" for developers who utilize multiple Anthropic-powered tools simultaneously. By providing a unified observability layer, it allows users to monitor their total burn rate across disparate platforms like Cursor, Windsurf, and Claude Code CLI. It integrates real-time Anthropic pricing models—including complex variables like Prompt Caching (Write/Read)—to provide high-precision financial reporting that would otherwise require manual calculation or multiple API dashboard checks.

Main Features

1. Automated Multi-Source Discovery: The tool utilizes a signature-based scanning engine that automatically detects and parses data directories for over nine major AI integrations. It targets specific file paths such as ~/.cursor/projects/, ~/.claude/projects/, and ~/.continue/sessions/. By parsing JSONL and session log files natively, it reconstructs conversation histories and token counts without requiring the user to manually input API keys or upload log files.

2. Granular Cost & Token Analytics Dashboard: The interface, powered by Chart.js, provides a comprehensive breakdown of spending. It tracks metrics across the entire Claude model family, including Opus (4.5/4.0), Sonnet (3.5/3.7/4.0), and Haiku (3.0/3.5/4.0). Features include daily cost charts, model-specific expenditure ratios, and animated stat cards. The dashboard also includes a "Monthly Projection" algorithm that calculates anticipated spending based on current velocity, helping developers stay within budget.

3. Advanced Usage Visualization (Heatmaps & Session Logs): To identify peak productivity or high-cost periods, the application generates GitHub-style contribution heatmaps and "Peak Hour" grids. This allows users to visualize temporal usage patterns. Furthermore, the "Session Log" feature provides an expandable, color-coded drill-down into every individual interaction, revealing exactly which prompt or tool (e.g., Aider, Cline, or Roo Code) was responsible for the highest token expenditure.

4. Privacy-Centric Native Architecture: Unlike many analytics tools that require cloud syncing or telemetry, Claude Usage Tracker is a "No-Cloud" solution. It is a native macOS application, not an Electron app, ensuring minimal resource footprint and high performance. All data extraction, calculation, and rendering happen locally, ensuring that sensitive codebase interactions and prompt histories never leave the user's hardware.

Problems Solved

1. Fragmented Usage Visibility: Developers using multiple tools often have no way of knowing their cumulative daily spend until they receive an Anthropic invoice. Claude Usage Tracker solves this "fragmentation problem" by merging usage data from CLI tools, IDE extensions, and standalone desktop apps into a single source of truth.

2. Hidden Costs of Prompt Caching: Anthropic's pricing involves complex tiers for "Cache Write" and "Cache Read" tokens. Manually calculating these costs is nearly impossible for the average user. This tracker automates the calculation of caching discounts and premiums, providing an accurate financial reflection of modern AI workflows.

3. Target Audience:

• AI-Native Software Engineers: Users who switch between Cursor for coding, Claude Code for CLI tasks, and Aider for pair programming.
• Engineering Leads & CTOs: Decision-makers who need to audit the cost-efficiency of different AI tools within their team's local environments.
• Freelance Developers: Individuals working on strict budgets who need to project their monthly API overhead.
• Privacy-Conscious Users: Developers who refuse to use third-party cloud trackers due to the sensitivity of their prompt data.

4. Use Cases:

• Budget Monitoring: Identifying if a specific project in Cursor is consuming more tokens than anticipated.
• Tool Comparison: Determining which AI assistant (e.g., Windsurf vs. Cline) is more efficient in token usage for similar tasks.
• Productivity Auditing: Using heatmaps to identify when the most intensive coding sessions occur.

Unique Advantages

1. Differentiation from Standard Dashboards: Standard API dashboards only show total API usage at the key level. Claude Usage Tracker differentiates itself by providing "App-Level" granularity. It can tell you specifically how much you spent within "Cline" versus how much was spent in "Claude Desktop," a distinction that the official Anthropic console cannot make.

2. Key Innovation: Local Session Reconstruction: The specific innovation lies in its ability to reverse-engineer local session formats. Many AI tools store data in obscure JSONL or SQLite formats; this tracker includes a specialized parser for each tool, enabling it to extract not just token counts, but also the context of the session (e.g., which project was being worked on) without an active internet connection.

3. Performance (Native vs. Electron): By leveraging a Swift-based native wrapper and shell scripts for data collection, the app avoids the memory bloat associated with Electron-based tools, making it an ideal background utility for developers who are already running resource-heavy IDEs.

Frequently Asked Questions (FAQ)

1. Is Claude Usage Tracker compatible with Windows or Linux? The primary distribution is a native macOS (.app) bundle designed for Apple Silicon and Intel. However, the core logic is built on Node.js (collect-usage.js). Users on Windows or Linux can run the tracker in "Browser Mode" by executing the script via Node and serving the dashboard over a local Python HTTP server.

2. Does this tool require my Anthropic API Key? No. Claude Usage Tracker does not require your API keys. It functions by reading the local log files and session data already stored on your hard drive by tools like Cursor or Claude Code. This ensures your credentials remain secure and never exposed to the application.

3. How does the tracker handle new Claude models like Sonnet 3.7 or Opus 4.5? The pricing engine is designed to be future-proof. It includes support for the entire upcoming 4.0 and 4.5 model families. Because the project is open-source (MIT License), pricing constants can be updated manually in the JavaScript logic, or automatically via repository updates as Anthropic adjusts its rates.

4. Can I export my usage data for tax or accounting purposes? Yes. The application includes built-in export and import functionality. This allows users to move their usage history between devices or export the aggregated data for further analysis in spreadsheet software, making it highly useful for professional expense tracking.