Token Saving Analyzer

Your customized plugin to cut token waste, up to 50% savings

36 followers

Your customized plugin to cut token waste, up to 50% savings

36 followers

Token Saving Analyzer builds you a plugin to cut token waste. It runs a deterministic profiler on your agentic coding logs (eg. Claude Code, Copilot, Cursor.) Wasteful patterns get identified and it generates a custom plugin that you can install to instantly cut waste and get up to 50% more use from your subscriptions.

Free

Launch tags:Productivity•Developer Tools•Tech

Launch Team

Viktor.comAn AI coworker that actually does the work

Promoted

Maker

📌

Hey Product Hunt, I built the Token Saving Analyzer because I wanted to get more out of my Claude, Codex, Cursor, and Copilot subscriptions. I don't have time to look at every session, so I built a tool that reads the logs and turns the report into a custom plugin. Install the plugin, save tokens. It's that easy. - Run `npx --yes agent-analyzer@latest run` - It scans supported local logs from Claude Code, Claude Desktop, Codex, OpenCode, Cursor, Kiro, and Google Antigravity - It writes a reviewable sanitized JSON report - The hosted report receives only that sanitized report, not raw transcripts The report shows the main waste patterns it found, a security receipt, benchmark-backed recommendations, and a generated remediation/plugin pack based on the same sanitized report. Then you can download your custom built plugin. It only contains the skills, commands, and 3rd party tool recommendations that would immediately help you based on the patterns found. The biggest design constraint was the privacy boundary. Raw agent logs can include prompts, local paths, command args, tool output, repo details, and secrets, so the public flow does not use a browser upload form. The command runs locally first and makes the upload boundary explicit. I would especially like feedback from people using Claude Code, Codex, Cursor, Kiro, OpenCode, or other coding agents every day: - Are the findings clear enough to act on? - Does the local-first flow feel trustworthy? - What waste pattern do you wish the report explained better? Thanks for taking a look.

Report

3mo ago

Maker

Thank you everybody who supported us on launch day! I hope the product helps you cut down on your token use. I'm also keen to get feedback from those of you who actually downloaded and installed the plugin that the tool generates (that's the main point of the product, after all!) Especially important would be to report if there are any problems with the skills, commands, or tools recommended? I'm going to roll an update to the tool so that it can analyze your usage Post-installation of the plugin to prove the savings.

Cheers,

Robert

Report

3mo ago

Forum Threads

p/token-saving-analyzer

•

3mo ago

Does anyone use Codegraph for token savings and effeciency?

Codegraph was recommended to me: https://github.com/colbymchenry/...
I've now run our benchmark on it and the results were negative: +6,094 estimated tokens, +4,046 tool-output tokens, +54.3% API-rate cost.
I think a lot of tools do a cool thing, but don't necessarily lead to better agentic coding or saved tokens.

https://analyzer.spec-kitty.ai now has the benchmark results for codeanalyzer built in but doesn't include it in the recommended list of token-saving tools.

p/token-saving-analyzer

•

3mo ago

What tips and tricks have you found to save tokens and make your coding sessions last longer?

I did some benchmarking, and the results are interesting.
https://analyzer.spec-kitty.ai/p...
Agent Analyzer guidance reduced measured token waste and API-rate cost in repeated controlled runs.

On the larger noisy repo, three fresh Agent Analyzer guided runs averaged 12,370 fewer estimated tokens, 12,698 fewer tool-output tokens, 504 fewer Claude output tokens while preserving the quality gate. Repriced at published Claude Sonnet 4.6 API rates, the repeated mean delta was -$0.059207, or 23.986% lower cost.
Some of the techniques and tools used are shown here:
https://analyzer.spec-kitty.ai/p...
We'll be launching the tool on PH tomorrow, hope you support us!

View all