This project addresses the lack of detailed visibility into Claude Code's execution, offering a desktop application that reconstructs and visualizes session activity from existing local logs. It targets Claude Code users who need to understand file operations, tool calls, and token consumption beyond the opaque summaries or noisy verbose outputs provided by the CLI. The primary benefit is restoring granular insight into Claude Code's behavior without altering its core functionality.

How It Works

The application reads raw session logs from the ~/.claude/ directory, reconstructing a detailed execution trace. It analyzes file reads, tool calls, token usage, context injections, and agent coordination, presenting this information in a structured, searchable, and visual interface. This approach is advantageous as it requires no modification to Claude Code itself, works with any existing session, and provides a clear middle ground between insufficient and overwhelming output.

Quick Start & Requirements

• Installation: Pre-built binaries are available for macOS (.dmg), Linux (.AppImage, .deb, .rpm, .pacman), and Windows (.exe). Docker (docker compose up) and standalone Node.js (pnpm install, pnpm standalone:build, node dist-standalone/index.cjs) options are also provided.
• Prerequisites: For development, Node.js 20+ and pnpm 10+ are required. General usage requires no API keys or configuration, only access to the ~/.claude/ log directory.
• Links: Docker/Standalone Deployment, Development

Highlighted Details

• Visible Context Reconstruction: Reverse-engineers and visualizes context injections, including file contents, skills, tool I/O, and token attribution per turn.
• Compaction Visualization: Detects and illustrates how Claude Code's context window fills, compresses, and refills over time.
• Custom Notification Triggers: Allows user-defined alerts for specific events (e.g., .env access, tool errors, high token usage) via regex matching.
• Rich Tool Call Inspector: Native rendering for read (syntax-highlighted code), edit (inline diffs), bash (output), and subagent (expandable execution trees) calls.
• Team & Subagent Visualization: Untangles and visualizes nested subagent execution trees and teammate messages, showing full team lifecycle.
• Command Palette & Cross-Session Search: Spotlight-style Cmd+K for project-wide session search.
• SSH Remote Sessions: Inspects Claude Code sessions on remote machines via SSH.
• Multi-Pane Layout: Enables side-by-side session comparison and flexible tab management.

Maintenance & Community

The project includes CONTRIBUTING.md and a Code of Conduct, indicating structured community engagement guidelines. No specific details on maintainers, sponsorships, or dedicated community channels (like Discord/Slack) are present in the README.

Licensing & Compatibility

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

Limitations & Caveats

Real-time updates in Docker or standalone Node.js deployments may exhibit slight delays compared to the Electron-based desktop application, as they rely on Server-Sent Events (SSE) over HTTP rather than native file system watchers.