Juejin Usage Reads Your Local AI Logs and Tells You Exactly How Dependent You Are

AI Tool Do you dare to see how many Tokens you burned in a month? Juejin Usage is here

Author: JavaDog程序狗 Tags: Juejin Community, AI Programming

How many lines of code did you actually let AI write for you in a month, how many Tokens did you burn, and how dependent are you on it?

I honestly never calculated it before. Until Xitu Juejin released an official tool Juejin Usage, which pulled out all the usage data of Claude and Codex on my computer with one click.

The moment I saw the dashboard, I fell silent.

I am Dog Brother, a coder in Qingdao. Today, no hype, no shade, I'll take you through this "usage-pulling" tool and share my real dashboard along the way.

🍊 What exactly is this thing

Official website: https://juejin.cn/aiusage/download

Juejin Usage is an open-source AI usage statistics tool from Xitu Juejin. It's not here to help you write code; it's here to "keep the books" for you.

When you normally use Agent tools like Claude Code or Codex to write code, every round of conversation and every Token actually has local logs. Juejin Usage automatically identifies the Agent tools you have installed, reads the local logs, and then organizes them into a usage dashboard that you can view, share, and even rank on a leaderboard.

It supports both desktop (macOS / Windows, install and use) and CLI forms. On first launch, the desktop version automatically registers Claude / Codex Hooks, syncs local usage, and the dashboard pops up on its own, directly showing you trends and model distribution.

👽 Plain language explanation It's like the "Screen Time" on your phone, but this time it's not counting how long you scrolled TikTok, it's counting how long you bossed AI around to write code and how many Tokens you burned.

🎯 What dirty laundry can it air about you

After installing and opening the dashboard, you can see these items, which sting a little:

Usage Trends: View AI call volume by day or week, instantly seeing which day you were frantically grinding (or frantically slacking off and letting AI grind for you).
Model Distribution: Which model are you actually favoring, Claude or Codex? The proportion is crystal clear.
Multi-device Merging: Usage from your office computer + home computer, merged after logging into your Juejin account, nowhere to hide.
AI Usage Leaderboard: Usage can even go onto the official Juejin leaderboard, comparing with developers across the network to see who is more "terminally AI-dependent".

🍪 A quick look at my dashboard

The first glance after I installed it, the numbers indeed exceeded expectations a bit. I'll report mine first; after you install it, you report yours too:

★ Replace the numbers below with your own dashboard's real data ★

Statistical period: Last 30 days (dashboard defaults to viewing local 90 days, can merge multiple devices after login)
Total call count: ____ times
Token consumption: ____ (input / output separated is fine too)
Most used model: ____ (e.g., Claude / Codex, accounting for ____%)
Leaderboard rank: Surpassing ____% of users across the network

After seeing this set of numbers, I realized I truly can't go a day without AI writing code. Sharing the numbers is just a side thing; more importantly, it's for review: which tasks are suitable to write yourself, which can be handed off to AI, and the data can more or less reveal some patterns.

🔒 Privacy: It really isn't eavesdropping on your chats

Many people's first reaction: Will this thing upload all my conversations with AI?

I specifically went through its privacy agreement. It operates on a local-first principle, mainly explaining the following things:

Local Collection: Only stores Token usage, model name, and source channel. Data defaults to staying on your local machine at ~/.ai-usage/.
Absolutely Does Not Collect: Conversation content, Prompt text, hostname, project name, API Key — it touches none of them.
Cloud Sync: Requires you to actively enable it yourself, and only reports desensitized Token counts + anonymous device identifiers.
Your Control: Turn off sync anytime, delete the local directory anytime to clear all records.

👽 Plain language explanation It records "you made AI do 100 tasks today", not "what tasks you made AI do". The ledger belongs to you, the content does not belong to it.

To put it plainly, Juejin Usage is more like an accountant, not someone here to listen to what you chatted about.

🛠️ How to install (step-by-step)

There are two installation methods, pick whichever suits you:

Official website: https://juejin.cn/aiusage/download

Method One: Desktop (most hassle-free)

Method Two: CLI (the romance of a programmer)

# Requires Node.js >= 20
npx @juejin-opensource/jusage start

🎯 Dog Brother's conclusion

My biggest feeling isn't "how great this tool is", but that I'm more inseparable from AI than I imagined.

For me, its real usefulness boils down to three things:

Reconciliation: You know exactly how much you spent on AI and how many Tokens you burned.
Review: Which type of work should be written by yourself, which should be tossed to AI — the data speaks.
Leaderboard Research: See which models developers across the network are favoring, don't work in isolation.

Do you dare to show your own AI usage? After installing, drop your total call count, most-used model, and leaderboard rank in the comments, and let's see who is truly "terminally AI-dependent" 👇