What is a 'good' cache hit rate for Claude Code?

Generally, a cache hit rate above 70% is considered excellent, while a rate below 50% suggests that your context is changing too frequently and may benefit from manual compaction.

How does this skill calculate API costs?

It uses model-specific pricing data for the Claude 3 and 4 families, differentiating between base input, output, cache writes, and cache reads to provide a high-accuracy estimate.

Where does the skill find session data?

The skill automatically scans the ~/.claude/projects/ directory for JSONL log files generated by the Claude Code CLI.

Can I analyze a specific project or session?

Yes, the skill supports command-line arguments to analyze a specific number of recent sessions or a direct path to a specific session.jsonl file.

Session Token Analytics

Name: Session Token Analytics
Author: SaintPepsi

bySaintPepsi

0•

分析与监控

Analyzes Claude Code session logs to optimize token usage, monitor cache performance, and provide detailed cost breakdowns.

Session Token Analytics is a specialized diagnostic tool for Claude Code users who need to monitor and optimize their resource consumption. By parsing JSONL logs from local project directories, the skill generates comprehensive reports on cache hit rates, context growth patterns, and estimated API costs across various Claude models. It helps developers identify when to use the compaction command, recognize inefficient tool usage patterns, and manage long-running coding sessions more cost-effectively by providing actionable efficiency recommendations based on real usage data.

主要功能

01Specific recommendations for reducing spend and improving session speed

02Automated cost estimation based on model-specific API pricing for Opus, Sonnet, and Haiku

030 GitHub stars

04Detailed breakdown of cache hits, cache creation, and uncached input tokens

05Cross-session comparison tables to track efficiency trends over time

06Detection of context growth curves and automatic compaction events

使用场景

01Optimizing context management to improve model response speed and reduce latency

02Benchmarking different Claude models to find the most cost-effective workflow

03Auditing API costs after intensive development or refactoring sessions

主要功能

01Specific recommendations for reducing spend and improving session speed

02Automated cost estimation based on model-specific API pricing for Opus, Sonnet, and Haiku

030 GitHub stars

04Detailed breakdown of cache hits, cache creation, and uncached input tokens

05Cross-session comparison tables to track efficiency trends over time

06Detection of context growth curves and automatic compaction events

使用场景

01Optimizing context management to improve model response speed and reduce latency

02Benchmarking different Claude models to find the most cost-effective workflow

03Auditing API costs after intensive development or refactoring sessions