HashMeterAi
100% Local & Private

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 — including unlimited local Ollama and cloud models from 10 API providers — unified into one clean dashboard, each in its own tab by model, with GLM, MiniMax, and Gemini counted under Claude Code. Your data never leaves your machine.

Get HashMeterAi How it works
Zeronetwork calls
9+AI tools unified
15 MBlight native app
No telemetry No accounts No cloud Read-only
HashMeterAi dashboard showing tokens processed, AI use time, estimated cost, calendar heatmap, and stats.

A privacy-first dashboard that reads the raw local transcripts your tools already write.

Runs on your machine

Every byte is processed locally. There is no backend, no cloud sync, and no remote API for usage data.

Reads metadata, never content

Token counts, timestamps, and model names — never your prompts, code, or AI replies.

Cannot phone home

Locked capability set. No telemetry, no analytics, no auto-update pings. The app simply cannot make network calls.

What you get

Honest numbers, beautifully simple

Your tools' built-in meters only count themselves. HashMeterAi reads the real local record and shows the true picture — across every tool, in one place.

The money number

Estimated dollar value of AI compute used at public API list rates — the stat everyone wants to screenshot.

Processed tokens

The truest measure of work the model actually did — real work, 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, computed from real session gaps in your transcripts.

Your AI persona

"Prompt Architect, Night-Shift Shipper" — generated from your real patterns, with every claim backed by a number.

Where you stand

An honest "top X%" read against a documented, fully offline benchmark — always shown with the number behind it.

Achievements

60 ranked trophies across Volume, Intensity, and Mastery — each a real threshold on your real-work tokens, with a true earn date and a tier-shaped medallion.

Calendar of activity

Your daily usage as a real heat-map calendar. Click any day for tokens, sessions, focus time, and top model.

One-click Share card

A clean branded image of your stats, persona, and three rarest trophies — pick a theme, then copy to clipboard or save as PNG. Brag, honestly.

How it works

Three steps to the truth

No accounts. No cloud. No setup. Open it and the numbers are already there.

1

Install in seconds

Grab the native app for macOS, Windows, or Linux. It runs entirely on your machine and never phones home — no account, no sign-up, nothing to configure.

2

It reads what's already there

HashMeterAi quietly reads the usage logs Claude, Codex, Kimi, Qwen CLI, HashCortx, and HashCerebrum already write to disk — plus GLM, MiniMax, and Gemini run through Claude Code. Only counts, timestamps, and model names. Never your prompts, code, or replies.

3

See the real you

One clean dashboard reveals your true cost, real-work tokens, focus time, AI persona, trophies, and activity calendar — unified across every tool. Then export the brag card and post it. Honestly.

Supported tools

One dashboard for every AI coder

Supported

Claude Code Anthropic
Supported
Codex OpenAI
Supported
Kimi Moonshot AI · by model
Supported
Qwen CLI Alibaba · Qwen Code CLI · by model
Supported
GLM Z.ai · counted under Claude
Via Claude
MiniMax counted under Claude
Via Claude
Gemini Google · counted under Claude
Via Claude
HashCortx Unlimited local Ollama + 10 cloud providers
Supported
HashCerebrum Local medical-research workbench
Supported

Not supported

Cursor Server-side data
Not supported
Copilot Server-side data
Not supported
Windsurf Server-side data
Not supported
Qwen desktop app Chromium IndexedDB · binary
Not supported
Antigravity Google · binary protobuf, no token counts
Not supported
Gemini CLI Telemetry off by default
Not supported

GLM, MiniMax, and Gemini have no usage files of their own — they run through Claude Code, so HashMeterAi reads them from Claude's local transcripts and counts them under the Claude tab, each keeping its own model name. Cursor, Copilot, and Windsurf keep usage data server-side; the Qwen desktop app stores chat in a Chromium IndexedDB, Antigravity in binary protobuf blobs with no token counts, and the Gemini CLI only logs tokens when telemetry is enabled (off by default) — so none of these can be measured locally.

Privacy is the whole point

Read a little.
Send nothing.

HashMeterAi is built to read only the usage metadata your tools already store on disk. It cannot connect to the internet, cannot read your prompts, and cannot touch your credentials.

Read the security policy
Zero network

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

Metadata only

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

Never secrets

Skips auth files, credentials, and keys entirely.

Read-only

It cannot modify any tool's data. Ever.

Inside the app

Built for clarity, not vanity

Persona
HashMeterAi Persona screen showing your AI persona with a top-percent standing and evidence-backed stats.
Achievements
HashMeterAi Trophies screen showing legendary, epic, rare, and common trophies organized by category.
Share card preview
The Architect

Built by someone who wanted the truth

Dr. Seif Hashish

Founder & Chief Architect

A clinical pharmacist and AI expert, Dr. Seif Hashish sits at the rare intersection of biomedical science and deep-learning research.

NGU University — Class of 2025 Clinical Pharmacy Expert Vibecoder
From the same developer

Part of the Hash ecosystem

HashMeterAi measures the usage of its sibling apps — both local-first, privacy-first, and built by the same hand.

HashCortX

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

View on GitHub

HashCerebrum

A local-first medical-research workbench with a 3D brain interface for searching, citing, and peer-reviewing research — all on your machine.

View on GitHub

Ready to see your real AI usage?

HashMeterAi is in active development. Build from source today, or watch the repo for the first native release.

View on GitHub Build from source
macOS
$
$
$
Windows
>
>
>
>
Linux
$
$
$

Requires Rust (stable) and Node 18+. Works on macOS, Windows, and Linux.