MaxForAI

MaxForAI / Tokenless

Public

One command to cut token usage by up to 50%+

agent agent-skills agentic-workflow agents anthropic
15
1
85% credibility
Found May 19, 2026 at 23 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
JavaScript
AI Summary

Tokenless is a local plugin for Claude Code that reduces AI costs by compressing the information sent to the AI while keeping all original data saved on your computer. When you read large files, run tests, or execute commands, Tokenless intercepts the output, saves the full version locally, and sends a compact summary to the AI instead. You can expand these summaries anytime you need the details. It also offers different response styles—chat mode for conversational replies or coding mode for dense technical output—so you control how verbose the AI responses are. Everything runs locally on your machine with no external services, and risky or failed outputs always pass through uncompressed.

How It Works

1
💡 You notice the costs adding up

As you use Claude Code for bigger projects, you realize every file read, test result, and command output keeps getting sent along, making each request more expensive.

2
📦 You find Tokenless

You discover a tool that promises to cut your AI costs by half while keeping all your work saved locally.

3
⚡ You set it up in seconds

With just a few simple commands, Tokenless is installed and ready to work alongside Claude Code.

4
🤖 Claude Code works as usual, but smarter

You keep coding normally, but now large outputs get compressed into summaries while the full details stay safe on your computer.

5
You choose how detailed you want responses
💬
Chat mode

Short, readable responses in plain language for everyday questions and discussions.

💻
Coding mode

Dense, technical output with abbreviations for efficient programming sessions.

6
🔍 You can always pull up the full details

When you need exact details from a compressed output, just ask to expand it and everything is there.

✅ You save money without losing anything

Your AI costs drop significantly while every piece of evidence from your work stays safely stored locally, ready whenever you need it.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 23 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 Tokenless?

Tokenless is a CLI tool that slashes your Claude Code bill by compressing what gets sent to the API. It works by intercepting tool outputs, file reads, and assistant responses, storing the originals locally, and replacing them with compact summaries. When you need the real data, you can expand specific sections on demand. The tool runs entirely on your machine, uses Claude Code hooks for integration, and offers three output profiles: chat for readable responses, coding for dense structured output, and off to disable everything.

Why is it gaining traction?

The hook is simple: real money. Their benchmarks show 47% fewer request tokens in vibe-coding sessions and 80% fewer response tokens in natural conversation. The tool handles the noisy stuff developers actually deal with: test logs, build output, git diffs, search results, and large file reads. The key insight is that raw evidence stays local while Claude gets a packet with anchors, snippets, and exact expansion commands when needed. No external LLM calls, no cloud summarization, just deterministic compression with a safety-first approach that passes through errors and risky outputs unchanged.

Who should use this?

Heavy Claude Code users who run long sessions or work with large codebases will see the biggest savings. Developers building React dashboards, editing multi-file projects, or running CI pipelines will benefit most from the tool output compression. If you're already using Claude Code casually, the overhead is minimal; if you're running it professionally, the token savings compound quickly. Teams doing high-stakes work in legal, financial, or security contexts will appreciate that raw artifacts remain locally available for exact review.

Verdict

Tokenless delivers on its core promise with solid benchmarks and a privacy-respecting architecture. The 15-star count and GitHub-only distribution reflect early-stage maturity, but the codebase includes comprehensive eval scripts and the MIT license lowers adoption risk. Try it if token costs are a concern; start with the chat profile for immediate savings without changing your workflow.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.