Developer AgentSeal released CodeBurn on April 13, 2026, a terminal dashboard providing real-time visibility into AI coding token consumption. The open-source tool gained 545 GitHub stars within 24 hours, addressing a critical pain point for developers using Claude Code and similar AI assistants.
Solving the AI Coding Cost Visibility Gap
CodeBurn reads session transcripts directly from disk to break down token spending by task type, tool, model, MCP server, and project. The tool requires no API keys, proxies, or wrappers—it analyzes local ~/.claude/sessions files to provide granular cost tracking.
The dashboard introduces a "one-shot success rate" metric that reveals when AI succeeds on the first attempt versus requiring multiple edit/test/fix retry cycles. This metric helps developers identify inefficient workflows and understand where AI coding assistants struggle.
Feature Set and Technical Implementation
CodeBurn delivers comprehensive cost observability through:
- Interactive TUI dashboard with real-time cost tracking across sessions
- 13 deterministic task categories: coding, debugging, testing, refactoring, documentation, and more
- macOS menu bar widget via SwiftBar for persistent monitoring without opening the terminal
- CSV/JSON export capabilities for further analysis and reporting
- Pricing data integration from LiteLLM with automatic caching
- Multi-dimensional breakdowns by task, tool, model, and project
Built in TypeScript, the tool targets developers seeking transparency in AI coding costs. The repository includes 27 commits on the main branch and covers topics including ai-coding, claude-code, cli, cost-tracking, developer-tools, observability, terminal-ui, and token-usage.
Rapid Organic Growth Signals Developer Demand
The tool's 545-star growth in under 24 hours reflects strong organic interest from developers facing opacity in AI coding costs. Without a Hacker News discussion, the GitHub traction suggests developers discovered CodeBurn through direct searches for cost tracking solutions or community recommendations.
AgentSeal describes the project as providing "interactive TUI dashboard for Claude Code and Codex cost observability," positioning it as essential infrastructure for teams managing AI coding budgets. The one-shot success rate metric offers particular value for understanding model performance patterns and optimizing prompting strategies.
Key Takeaways
- CodeBurn gained 545 GitHub stars within 24 hours of its April 13, 2026 release
- The tool provides real-time token cost tracking for AI coding assistants like Claude Code without requiring API keys or proxies
- A unique "one-shot success rate" metric reveals when AI succeeds immediately versus needing multiple retry cycles
- CodeBurn reads local ~/.claude/sessions files and categorizes spending across 13 task types, tools, models, and projects
- Built in TypeScript with macOS menu bar widget support and CSV/JSON export capabilities