oanhduong

oanhduong / token-ninja

Public

token-ninja routes deterministic shell commands locally — zero LLM calls, ~19µs latency. Works silently inside AI tools via MCP.

ai-tools antigravity claude-code codex copilot
15
1
100% credibility
Found Apr 20, 2026 at 15 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Token-ninja intercepts common shell commands in AI coding assistants like Claude Code and runs them locally to save tokens, latency, and costs while preserving safety.

How It Works

1
🔍 Discover token-ninja

You hear about a helpful sidekick for your AI coding helper that runs everyday commands super fast without wasting your AI credits.

2
📦 Get it with one click

You add it to your computer in seconds, and it quietly sets everything up for your favorite AI tools.

3
💬 Chat as usual

Open your AI coding buddy and start asking about your project, like checking file changes or running tests.

4
⚡ Watch the magic

Simple commands appear instantly with full colors, just like your own terminal, and a tiny note shows how many AI credits you saved.

5
📊 See your savings grow

Check a quick report anytime to see total credits saved and which commands helped the most.

🎉 Work faster, spend less

Your coding sessions feel snappier, your AI focuses on the smart stuff, and your wallet stays happy.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 15 to 15 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is token-ninja?

Token-ninja intercepts shell commands like "git status" or "npm test" in AI coding tools such as Claude Code, Cursor, or Aider, routing them locally for zero LLM calls and ~19µs latency. Built in TypeScript, it uses MCP for seamless integration, preserving ANSI colors and adding a subtle token-savings footer—conversational prompts pass through unchanged. Developers save tokens and time on deterministic tasks across git, docker, npm, and 46 other domains.

Why is it gaining traction?

It stands out with 736 rules covering real workflows (e.g., "show recent commits" → git log), double safety checks blocking rm -rf or sudo, and one-line install that auto-hooks Claude Code's UserPromptSubmit plus MCP for others—no prefixes or new habits needed. Unlike token ninja lite bb4124 alternatives or token ninja bb4124/token ninja bb86 variants, it benchmarks at 92% test coverage, runs commands inside AI sessions silently, and estimates savings via "ninja stats". Low-latency local execution beats LLM roundtrips for everyday calls and commands.

Who should use this?

AI-assisted coders chatting "docker ps" or "run the tests" in Claude Code, Cursor, or Continue—full-stack devs, DevOps folks managing k8s/helm, or anyone burning tokens on git/npm/docker queries. Ideal for high-volume AI sessions where 400+ tokens per status check adds up, but skip if you avoid MCP tools.

Verdict

Grab it if you use Claude Code daily—solid docs, 234 passing tests, and ninja setup/uninstall make it low-risk despite 15 stars and 1.0% credibility score. Early but battle-tested; extend with custom YAML rules for your stack.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.