Sarins Lab
- Updated June 2, 2026
- sarins-lab/code-optimization-plugin
Sarins Lab code-optimization-plugin is a Claude Code marketplace plugin that analyzes session JSONL transcripts so developers can see token costs, waste patterns, and improvement trends without manual tracking.
About
The Sarins Lab code-optimization marketplace plugin installs a Claude Code extension that mines your local JSONL session logs for token usage, waste patterns, and week-over-week trends. Developers reach for it when agent sessions feel expensive or repetitive and they want concrete, prioritized fixes instead of guessing. Tools like cost_trend and suggest_optimizations turn transcript data into retrospectives you can act on in the next session.
- Six MCP tools: cost_summary, cost_trend, suggest_optimizations, file_read_analysis, top_repeated_tasks, summarise
- Rule-based before/after health verdicts with persisted ~/.claude/cost-analysis-state.json
- Surfaces redundant file reads, large-context turns, and retry loops from JSONL
- No manual tracking; analyzes actual Claude Code session transcripts
Sarins Lab by the numbers
- Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add sarins-lab/code-optimization-pluginAdd your badge
Show developers this marketplace is listed on Skillselion. Paste this into your README.
| Last updated | June 2, 2026 |
|---|---|
| Repository | sarins-lab/code-optimization-plugin ↗ |
How do you know which Claude Code habits burn tokens and what to fix first?
Measure Claude Code token spend from JSONL transcripts and get prioritized fixes for redundant reads, fix loops, and repeated tasks.
Who is it for?
Teams running heavy Claude Code sessions who want data-backed cost and efficiency retrospectives from local transcripts.
Skip if: Developers who do not use Claude Code or who cannot access local session JSONL logs.
What you get
You receive token breakdowns, trend deltas, prioritized waste findings, and optional baseline comparisons stored in local plugin state.
Plugins in this marketplace
1 plugin - install individually after you add the marketplace.
Recommended Marketplaces
FAQ
Where does usage data come from?
The plugin reads Claude Code session JSONL transcripts on your machine; it does not require manual logging.
What does summarise do on first run?
It captures a baseline snapshot and returns BASELINE_SET; later runs compare metrics against that state.
How many tools ship in the plugin?
The marketplace describes a cost-analysis MCP server with six tools including cost_summary and suggest_optimizations.