AI News HubLIVE
站内改写5 min read

Show HN: HashMeterAi – Private AI Token Real Usage Meter for All Models

HashMeterAi is a local-first, private usage meter for AI coding tools. It tracks token usage across multiple tools like Claude Code, Codex, Kimi, Qwen CLI, and others, providing a unified dashboard with metrics like cost, processed tokens, focus time, and AI persona. It is 100% offline and privacy-focused, never sending data. Supports several tools via local transcripts.

SourceHacker News AIAuthor: Hash7777

Notifications You must be signed in to change notification settings

Fork 0

Star 6

BranchesTags

Open more actions menu

Folders and files

NameName

Last commit message

Last commit date

Latest commit

History

131 Commits

131 Commits

src-tauri

src-tauri

src

src

.gitignore

.gitignore

LICENSE

LICENSE

README.md

README.md

SECURITY.md

SECURITY.md

package.json

package.json

Repository files navigation

See how much AI you really use.

The honest, local-first usage meter for AI coding tools. Claude Code, Codex, Kimi, Qwen CLI, HashCortx, HashCerebrum — unified into one clean dashboard with usage based trophies :)

Your tools' built-in meters skip sessions, miss whole days, and only count themselves. HashMeterAi reads the raw local transcripts every tool already writes, and shows you the true picture — across all of them, in one place. 100% offline. Nothing leaves your machine.

What you get

The money number. Your estimated dollar value of AI compute used at public API rates — the stat everyone wants to screenshot.

Processed tokens. The truest measure of work the model actually did (not inflated by cached re-reads), with the "approximately N times The Lord of the Rings" yardstick.

Average focus time. Honest active-time-with-AI per day.

Your AI persona. For example, "Deep Diver, Night Owl" — generated from your real patterns, and every claim shows the number behind it. No vanity, no lies.

Where you stand. An honest "top X% of AI developers" read, computed against a documented, fully offline usage benchmark modeled from public 2025–26 figures — always shown with the number behind it, never a made-up rank.

Achievements. Sixty ranked trophies across Volume, Intensity, and Mastery (twenty each) — from First Steps to the Billion-token clubs, Night Owl to Whale. Every token trophy counts processed (real-work) tokens, not the inflated billed footprint, and shows its true earn date on a tier-shaped medallion. Your three rarest headline your Share card.

Calendar of activity. Your daily usage as a real calendar.

One-click Share card. A clean branded image of your stats and persona to post. Brag, honestly.

Privacy is the whole point

HashMeterAi is built to read a little and send nothing:

Guarantee

Zero network No telemetry, no analytics, no auto-update pings. It cannot phone home.

Metadata only Reads token counts, timestamps, and model names — never your prompts, code, or replies.

Never touches secrets Skips auth files, credentials, and keys entirely.

Read-only It cannot modify any tool's data. Ever.

100% local Everything is computed on your machine. Your usage is yours.

The only thing that ever leaves is a brag-card image that you choose to export. See SECURITY.md.

Supported tools

Tool Status

Claude Code Supported. Also counts GLM (Z.ai), MiniMax, and Gemini when they're run through Claude Code via an Anthropic-compatible endpoint — those turns stay on the Claude tab and keep their own model name in the Models breakdown.

Codex (OpenAI) Supported

Kimi Supported (classified by model)

Qwen CLI Supported — reads the Qwen Code CLI (~/.qwen), classified by model

HashCortx Supported (records real per-response token counts)

HashCerebrum Supported (records real per-response token counts)

Qwen desktop app Not supported — chat lives in a Chromium IndexedDB (binary LevelDB); no readable per-message token data

Antigravity (Google) Not supported — conversations are stored as binary protobuf blobs with no token-count fields recorded locally

Gemini CLI Not supported — token counts are only written when telemetry logging is enabled (off by default)

Cursor, Copilot, Windsurf Not supported — usage is server-side; no local data to read

Models. Model names and public list-price rates are recognized for the Claude, GPT/Codex, Kimi, Qwen, GLM, MiniMax, and Gemini families — so even a non-Anthropic model run through Claude Code is named and costed correctly, never lumped under a generic label.

Adding a tool is one Rust file — the architecture is a pluggable source adapter. PRs welcome.

Screenshots

Install

Download the installer for your OS from the Releases page, or build from source (below).

Why your OS shows a warning. HashMeterAi is a free, open-source project and ships unsigned — code-signing certificates (an Apple Developer ID, a Windows EV certificate) are paid, ongoing costs. The warning means "not signed," not "unsafe": there's no network access, and every line of source is in this repo for you to read or build yourself. Each warning below is a one-time step.

macOS — .dmg

Open the .dmg and drag HashMeterAi to Applications.

On first launch macOS says "HashMeterAi can't be opened because Apple cannot check it for malicious software." Get past it once, either way:

Right-click (or Control-click) the app → Open → Open, or

System Settings → Privacy & Security, scroll to the HashMeterAi message → Open Anyway.

If instead you see "HashMeterAi is damaged and can't be opened" (quarantine, common on Apple Silicon), clear the quarantine flag in Terminal, then reopen:

xattr -dr com.apple.quarantine /Applications/HashMeterAi.app

Windows — .msi / .exe

Run the installer. Windows SmartScreen shows "Windows protected your PC — unknown publisher."

Click More info → Run anyway (one time only).

If your browser flags the download, choose Keep.

Linux — .AppImage or .deb

Linux doesn't block unsigned apps the way macOS and Windows do, so there's no "unknown developer" prompt.

AppImage — make it executable and run:

chmod +x HashMeterAi_*.AppImage ./HashMeterAi_*.AppImage

If it won't start, install FUSE: sudo apt install libfuse2 (Debian/Ubuntu) or your distro's equivalent.

.deb (Debian/Ubuntu):

sudo dpkg -i HashMeterAi_*.deb sudo apt-get install -f # pull in any missing dependencies

Build from source

git clone https://github.com/Hash-7777/HashMeterAi.git cd HashMeterAi npm install npm run tauri dev # run it npm run tauri build # produce a native bundle in src-tauri/target/release/bundle/

Requires Rust (stable) and Node 18+. Building from source skips every OS warning above — the app is local to your machine.

Tech

Tauri v2 + Rust + vanilla JavaScript — no Electron, no bundler. One small native binary (target under 15 MB). All parsing and aggregation in Rust (parallel and incremental, so refresh is instant). The UI is plain HTML, CSS, and JS.

The Hash ecosystem

HashMeterAi is one of three local-first, privacy-first apps by the same developer — no cloud, no telemetry, your data stays on your machine:

App What it is

HashMeterAi (this app) See how much AI you really use — the honest local usage meter for AI coding tools.

HashCortX The local-first AI workspace — eleven modes, ten providers, zero telemetry.

HashCerebrum A local-first medical-research workbench with a 3D brain interface for searching, citing, and peer-reviewing research.

HashCortX and HashCerebrum write a local token-usage log that HashMeterAi reads, so your usage across the whole ecosystem is measured accurately in one place.

Contributing

Issues and PRs welcome — especially new source adapters. Keep it honest, local, and light.

Disclaimer

Estimates, not bills. The cost figure is the value of compute at public API list prices — on a subscription you don't pay per token. Token counts are read from each tool's local files and may differ from a provider's official dashboard.

Not a live ranking. The "top X%" standing is a percentile against a static, modeled 2025–26 usage benchmark computed entirely on your machine — it is labeled as a modeled benchmark, not a real-time leaderboard.

Trademarks. All product names, logos, and brands are the property of their respective owners and are used for identification only. HashMeterAi is independent and is not affiliated with, endorsed by, or sponsored by Anthropic, OpenAI, Google, Alibaba, Moonshot AI, Z.ai, MiniMax, or any other tool named.

No warranty. HashMeterAi is provided "as is", without warranty of any kind, as set out in the license.

License

Apache-2.0, Copyright 2026 Seif Hashish (Hash-7777). Provided "as is", without warranty of any kind.

The dollar figure is the value of compute at public API list prices — on a subscription you don't pay per token. The "top X%" standing is a percentile against a static benchmark curve modeled from public 2025–26 AI-coding usage data (≈$6/dev-day on Claude Code, agentic tasks of 1–3.5M tokens, power users at millions/day) — computed entirely on your machine, shown with its underlying number, and labeled as a modeled benchmark, not a live ranking. Honest by design.

About

The honest, local-first usage meter for AI coding tools. With full privacy and with no server or reading your chats.

hashmeterai.com

Topics

api

rust

ai

tokens

codex

ai-agents

ai-model

ai-tools

ollama

token-usage

qwen

claude-code

coding-agent

hashcortx

Resources

Readme

License

Apache-2.0 license

Security policy

Security policy

Uh oh!

There was an error while loading. Please reload this page.

Activity

Stars

6 stars

Watchers

0 watching

Forks

0 forks

Report repository

Releases 2

HashMeterAi V1.1

Latest

Jun 15, 2026

+ 1 release

Packages 0

Uh oh!

There was an error while loading. Please reload this page.

Contributors

Uh oh!

There was an error while loading. Please reload this page.

Languages

Rust 48.2%

JavaScript 30.9%

CSS 15.2%

HTML 5.7%