Show HN: Claude Code Session Trace/Browse Tool (Python)

What Happened

A developer has released an open-source Python tool that provides a terminal-based interface for analyzing Claude Code sessions stored on disk. The tool allows users to browse through session traces, inspect which files were accessed, which tools were invoked, and where token consumption occurred during Claude interactions. It addresses a practical gap in the current Claude Code ecosystem: while sessions are logged locally, there has been no simple way to visualize or audit that data in a structured manner.

Why It Matters

This tool highlights a broader trend in AI-assisted development: the growing need for observability and cost transparency. As developers increasingly rely on Claude Code and similar agents for complex, multi-step tasks, understanding how tokens are spent becomes critical—not just for budgeting, but for debugging inefficient workflows. A session that burns thousands of tokens on repeated file reads or unnecessary tool calls is a session that could be optimized.

The tool’s focus on local session analysis is also significant. Many AI development tools push telemetry to cloud dashboards, raising privacy and latency concerns. By keeping analysis on-device and in a TUI, this approach aligns with the preferences of security-conscious developers and those working in air-gapped environments. It also enables offline debugging, which is essential when network access is limited or when working with sensitive codebases.

Implications for AI Practitioners

For developers using Claude Code daily, this tool offers a concrete way to audit agent behavior. Instead of guessing why a session felt slow or expensive, practitioners can now pinpoint exactly which file reads or tool invocations consumed the most tokens. This granularity supports better prompt engineering—for example, learning to avoid patterns that trigger unnecessary context loading.

The tool also sets a precedent for the broader AI tooling ecosystem. As more agents adopt session logging, the demand for local, open-source analysis tools will likely grow. This mirrors the early days of web analytics, where raw log files gave way to sophisticated dashboards. We may soon see similar tools for other AI coding assistants, or integrated features within Claude Code itself.

However, the tool’s current scope is limited to post-hoc analysis. Real-time monitoring or proactive alerts about token burn would be a natural next step. Additionally, the tool’s utility depends on the quality and completeness of Claude’s session logs—if the logging format changes, the tool will need updates.

Key Takeaways

• A new open-source Python TUI enables developers to analyze local Claude Code session traces for token usage, file access, and tool invocation patterns.
• The tool addresses a critical gap in AI development observability, helping practitioners debug costly or inefficient agent workflows.
• Local, offline analysis aligns with privacy and security needs, and may become a standard feature in AI coding tooling.
• Future improvements could include real-time monitoring and broader compatibility with other AI agent session formats.