GitHub's CodeBurn: Track AI Coding Tokens with Dashboard

Overview

According to GitHub Trending, CodeBurn is a new tool from developer AgentSeal that helps track AI coding token usage for services like Claude Code and Codex. It provides an interactive terminal-based dashboard to monitor costs by task type, model, and more, using data from local session files without needing API keys. Released recently on GitHub, it's designed for developers working with AI-assisted coding.

Why It Matters

For developers building AI automation, tracking token usage is essential to control costs, especially with models like those from OpenAI or Anthropic. CodeBurn lets you see exactly where tokens are spent—on successful runs versus retries—helping optimize workflows and avoid unexpected bills. If you're using AI in web development, this tool can highlight inefficiencies in your scripts, making it easier to refine prompts or select cheaper models. My take: it's a solid addition for anyone integrating AI, as it promotes better resource management without adding complexity to your stack.

On the technical side, CodeBurn pulls data directly from directories like ~/.claude/projects/ for Claude or ~/.codex/sessions/ for Codex, analyzing transcripts to generate reports. This approach avoids wrappers or proxies, keeping things lightweight. Commands are straightforward: run codeburn for a dashboard of the last seven days, or codeburn today for daily stats. It supports multiple providers via a plugin system, so you can toggle between Claude and Codex with a simple flag, like codeburn report --provider claude. The dashboard includes features like gradient charts and keyboard navigation, built with Node.js, which fits seamlessly into my own toolkit of React and Next.js projects.

Technical Details

CodeBurn's architecture relies on Node.js 20 or higher, reading files from specific paths to aggregate data on token usage. For instance, it calculates one-shot success rates by parsing session logs, showing how often AI completes tasks without edits. Installation is simple: use

with npm install -g codeburn, or run it via npx codeburn for quick tests. Once set up, the TUI dashboard displays responsive panels that update in real-time, with exports available in CSV or JSON formats using commands like codeburn export -f json.

Key trade-offs include its dependency on local file structures, which means it only works if your AI tools save sessions correctly—something that's reliable for Claude and Codex but might limit broader adoption. The provider system is extensible, planning support for tools like Pi or OpenCode, but currently focuses on established ones. In practice, this means you get detailed breakdowns, such as token counts per project, without overwhelming your system resources. For AI automation in Rails or Python scripts, integrating this could involve scripting around its outputs to automate cost alerts.

Pros and Cons

The main advantage is cost observability: it gives precise insights into AI usage, helping developers like me budget for token-heavy tasks in web apps. For example, in a React project with AI-generated components, CodeBurn could flag high-retry activities, saving money over time. Another plus is its no-fuss setup—no API keys needed—which enhances privacy and ease of use.

On the downside, it's limited to specific AI tools right now, so if you're using something else, you'll have to wait for plugins. Also, as a TUI app, it might not appeal to everyone who prefers graphical interfaces, and parsing local files could introduce errors if directories aren't standard. Overall, I recommend it for Node.js users monitoring AI costs, but it's not a full solution for non-terminal workflows.

FAQs

What is CodeBurn exactly? It's an open-source dashboard tool, available at

, that tracks AI coding token expenses by analyzing local session data from tools like Claude and Codex.

How do I get started with CodeBurn? Install it globally via

using npm install -g codeburn, then run codeburn to launch the dashboard. It requires Node.js 20+ and access to your AI session directories.

Is CodeBurn suitable for all AI models? It primarily supports Claude and Codex, with plans for more providers. If you're using other models, check for compatibility through its plugin system, but it's best for the supported ones to ensure accurate tracking.

---

📖 Related articles

• Lean-ctx: Ottimizzatore Ibrido Riduce Consumo Token LLM del 89-99%
• Rust rivoluziona Claude Code: Avvio 2.5x più rapido e volume ridotto del 97%
• GitHub: Claude e Obsidian per un vault di conoscenza persistente