Developers juggling multiple AI coding assistants now have a unified way to monitor their token consumption and AI fluency. The newly released AI Engineering Fluency extension, previously known as the GitHub Copilot Token Tracker, provides real-time tracking for a broad ecosystem of tools including GitHub Copilot, Claude Code, Gemini CLI, Continue, OpenCode, Crush, Pi CLI, and Mistral Vibe. All data is read from local session logs, ensuring privacy unless the user opts into cloud sync.

Extensive Editor and Fork Support

The extension works with all Chromium-based VS Code forks, explicitly supporting VS Code, VSCodium, Cursor, Windsurf, Trae, Kiro, and Void. For JetBrains users, it integrates into IntelliJ IDEA, Rider, PyCharm, WebStorm, GoLand, RubyMine, and CLion by reusing the same webview bundles found in the VS Code extension. Visual Studio 2022+ users can track usage by reading Copilot Chat session files directly. Additionally, a CLI tool allows for terminal-based stats via npx @rajbos/ai-engineering-fluency stats, catering to developers who prefer command-line interfaces.

GitHub Copilot App Canvas and CLI Integrations

A major new feature is the GitHub Copilot app canvas, a side-panel interface that displays token usage, cost, sessions, and fluency scores within the Copilot app. This canvas is powered by the local CLI and ships as the ai-fluency-canvas plugin. For terminal users, the tool offers an Oh-My-Posh segment that displays today's and 30-day token usage directly in the prompt. There is also an experimental GitHub Copilot CLI Statusline that combines context tokens, session duration, line changes, and daily totals into one compact line, configured via the settings.json file.

Privacy-First Architecture and Team Sharing

The privacy-first design ensures data stays local by default. When using the GitHub Copilot app canvas, data is only sent to the model if the user asks an agent about their stats, functioning like any other tool result. For teams, the project includes a self-hosted sharing server that runs on SQLite and Docker, requiring no Azure account. Team members can share aggregated usage data through a lightweight API server, with authentication handled via existing GitHub sessions.

Key Takeaways

  • Supports a wide range of AI tools beyond Copilot, including Claude Code, Gemini CLI, Continue, and OpenCode.
  • Works across all Chromium-based VS Code forks, JetBrains IDEs, Visual Studio 2022+, and the CLI.
  • Features a new GitHub Copilot app canvas for in-app usage visualization.
  • Integrates with terminal workflows via Oh-My-Posh segments and an experimental Copilot CLI statusline.

The Bottom Line

This tool finally solves the fragmentation problem of tracking AI costs across different environments and providers, making it an essential utility for any developer serious about managing their AI engineering budget.