Summary

Claude Code Trace addresses the need for real-time inspection of Claude Code session logs. It provides developers with a tool to browse, tail, and search logs, rendering conversations with expandable tool calls and token counts. The application offers a significant benefit by improving the debugging and understanding of complex AI-generated code sessions across multiple platforms.

How It Works

Built using Tauri v2 for a Rust backend and React for the frontend, the application processes session logs from ~/.claude/projects/ JSONL files. It renders these logs as scrollable conversations, allowing users to expand tool calls and view token counts. A key advantage is its support for the Model Context Protocol (MCP), which automatically detects and displays tool calls with human-friendly names from various supported servers like chrome-devtools and figma.

Quick Start & Requirements

Installation options include building from source (Rust 1.77+, Node.js 18+ required), downloading pre-built binaries for macOS, Linux, and Windows, or running via Docker for web mode. Building from source involves cloning the repository and running ./script/install.sh or npm install followed by npm run tauri dev. Pre-built binaries offer direct execution. Docker requires docker build and docker run. Specific Linux development libraries (libwebkit2gtk-4.1-dev, etc.) and macOS Xcode Command Line Tools are prerequisites for building. WebView2 is pre-installed on Windows.

Highlighted Details

• Real-time log tailing and searching capabilities.
• Detailed rendering of conversations with expandable tool calls and token counts.
• Automatic detection and human-friendly display of Model Context Protocol (MCP) tool calls.
• Cross-platform support: native desktop app (macOS, Linux, Windows), web app, and TUI.
• Comprehensive keyboard shortcuts for efficient navigation and interaction within the application.

Maintenance & Community

The project welcomes contributions via bug reports, feature requests, and pull requests. Specific details regarding maintainers, sponsorships, roadmaps, or dedicated community channels (like Discord or Slack) are not explicitly detailed in the README.

Licensing & Compatibility

The project is licensed under the MIT license. This permissive license generally allows for commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

The macOS pre-built application is unsigned and requires manual quarantine attribute removal (xattr -cr). The Terminal UI (TUI) is noted as functional but has some UX rough edges, with contributions actively sought to improve it.