io.github.TorutDev/turntax·AI & LLMs

turntax

Per-turn token cost of AGENTS.md, CLAUDE.md and other agent instruction files, with safe fixes.

Stars
0
Version
0.6.1
License
MIT
Review
Passed
Source Website npm
Language
TypeScript
Last commit
2h ago
Forks
0
Open issues
0
Registry update
Oct 16, 2026
agents-mdclaude-codeclicodexcursorgemini-cligithub-copilotllmtokenswindsurf

Connect

npm
Terminal
$claude mcp add turntax \  -- npx -y [email protected]

Environment variables

Nothing to configure — it works as is.

Security review

Review passed

Reviewed 33m ago.

metadata
scanned
packages
1 checked

No findings.

Tools

This server runs on your machine, so its tool list is not fetched before install — the README below describes what it offers.

README

TorutDev/turntax/plugins/turntax/README.md

turntax plugin for Claude Code

Measures how many tokens your agent instruction files (CLAUDE.md, .claude/rules, skills, subagents, AGENTS.md and the rule files of 15 other coding agents) add to every turn, and previews safe fixes.

What it adds

  • The turntax MCP server with two read-only tools: audit (the per-turn report for a folder) and fix_preview (the diff turntax --fix would make; nothing is written).
  • The /turntax:context-tax [folder] command, which runs both tools for the current project and summarises the largest per-turn costs. It applies fixes only when you ask for it explicitly.

What it runs and sends

The server runs npx -y [email protected] mcp on your machine. npx downloads the turntax package from npm on first use. The tools read instruction files and MCP configs in the folder you audit. They do not start the MCP servers they find, do not call any vendor API, and send nothing over the network.

Install

/plugin marketplace add TorutDev/turntax
/plugin install turntax@turntax

Source, CLI and documentation: https://github.com/TorutDev/turntax (MIT, TORUT LTD).

From the server's repository — third-party content, read it before running anything it suggests.

More in AI & LLMs

humanizer-ru

io.github.ilyautov

Следы нейросети в русском тексте: балл чистоты 0-100, запреты и маркеры с позициями, факт-замок.

Neural Draft

io.github.vbalagovic

Neural Draft MCP — CMS, blog, bookings, galleries, commerce. 38 tools.

mcp-contact-classifier

com.mambabuilt

Classify a job title into department and seniority, and optionally verify the position.

mcp-agent-accessibility-auditor

com.mambabuilt

Audit whether an AI agent can read a site: llms.txt, robots AI rules, structured data.

Token Lens

io.github.duttkunal101-lab

Count and cut tokens in prompts, chats and agents for any model. Runs locally, collects no data.

prompt-leak-scanner

io.github.GuardBee

Catches leaked credentials and PII in outbound LLM prompts

Listing from the official MCP Registry · repository facts from GitHub, 1h ago.