AI Fluency Tracker Brings Transparency to Copilot Usage

Overview

Developers seeking greater visibility into their artificial intelligence tool consumption now have a robust, open-source solution: the AI Engineering Fluency Tracker. Originally launched as the "GitHub Copilot Token Tracker," this project has expanded significantly to become a comprehensive observability platform for modern software engineering workflows. The tool provides real-time insights into token usage, fluency scores, and environmental impact across a wide array of integrated development environments and command-line interfaces.

By reading directly from local session logs, the tracker ensures that sensitive code data remains on the developer’s machine by default. This privacy-first architecture addresses growing concerns in the industry regarding telemetry and data sovereignty. Users who choose to enable cloud sync or query AI agents for statistics do so explicitly, maintaining full control over when and how their usage data is transmitted or displayed.

Platform Support and Installation

The tracker’s primary value proposition lies in its extensive compatibility with the fragmented landscape of AI-assisted coding tools. It is not limited to a single vendor’s ecosystem but instead aggregates data from multiple providers. This unified approach allows teams and individuals to gain a holistic view of their AI footprint, regardless of which specific tool they are using for a given task.

Integrated Development Environments

For developers working within graphical IDEs, dedicated extensions provide seamless integration. The VS Code Extension is available for the Chromium-based versions of VS Code, including stable, insider, and exploration builds. It displays real-time token usage in the status bar and offers a detailed fluency score dashboard. This extension is also compatible with various forks and derivatives such as VSCodium, Cursor, Windsurf, Trae, and Kiro.

JetBrains users are supported through a lightweight Kotlin/JCEF plugin that brings similar dashboard capabilities to IntelliJ IDEA, PyCharm, Rider, GoLand, and other JetBrains products. Meanwhile, Visual Studio 2022 and later versions are covered by a dedicated extension that reads Copilot Chat session files directly. These tools collectively ensure that IDE-centric developers have immediate access to their consumption metrics without leaving their primary workspace.

Command-Line and Terminal Integration

Recognizing that many AI engineers operate primarily in terminal environments, the project provides a versatile CLI tool powered by Node.js. Developers can run commands like npx @rajbos/ai-engineering-fluency stats to retrieve usage statistics, fluency scores, and even estimated environmental impact. This flexibility extends to Oh-My-Posh segments, allowing users to display daily and thirty-day token usage directly within their terminal prompt.

Furthermore, experimental support exists for the GitHub Copilot CLI statusline. By configuring a custom command in the user settings, developers can embed compact usage summaries—including context tokens, session duration, and line changes—into their command-line interface. This feature caters to power users who prefer minimal, information-dense terminals.

Privacy and Data Sovereignty

A critical distinction of this tracker is its local-first design philosophy. Unlike many cloud-native analytics platforms that require continuous data uploads, this tool processes logs stored locally on the user’s device. This approach minimizes the attack surface and protects intellectual property. The documentation explicitly states that nothing leaves the machine unless the user opts into cloud sync or interacts with an AI agent using their statistics as context.

Even when AI agents are used to interpret these numbers, the data is treated as standard tool input rather than automatically transmitted to telemetry servers. This transparency is crucial for enterprise environments where compliance and data governance are paramount. The optional self-hosted sharing server component further enhances this privacy model by allowing teams to share aggregated usage data across an organization without relying on third-party cloud services like Azure.

Expanding Beyond Copilot

While the tool began with a focus on GitHub Copilot, its scope has broadened to include a diverse range of AI engineering tools. The current list of supported applications includes:

• Claude Code and Claude Desktop Cowork (Anthropic)
• Gemini CLI (Google)
• Mistral Vibe
• Pi CLI (pi.dev)
• Kiro IDE and Kiro CLI (AWS)
• Continue and OpenCode

This expansion reflects the reality that modern developers often juggle multiple AI assistants. By consolidating metrics from Claude, Gemini, and other tools alongside Copilot, the tracker provides a singular pane of glass for AI fluency. This unified view is increasingly valuable as organizations attempt to manage costs and optimize workflows across heterogeneous AI stacks.

Impact on Developer Workflow

The introduction of such granular visibility tools signals a maturation in the AI software engineering market. As AI assistance becomes ubiquitous, the need for cost management and performance optimization grows in parallel. Fluency scores, which likely measure the efficiency and relevance of AI interactions, help developers refine their prompting strategies. Over time, this feedback loop can lead to more effective use of AI resources and reduced operational costs.

For technical leads and engineering managers, the ability to view team-wide usage patterns through self-hosted dashboards offers significant strategic advantages. It enables better budgeting for API calls and helps identify opportunities for training or process improvement. Ultimately, the AI Engineering Fluency Tracker represents a step toward more responsible and transparent adoption of AI technologies in professional development environments.