HashMeterAi is an "honest, local-first" usage statistics tool for AI programming tools, with a core philosophy of complete privacy, no servers, and never reading your chat content. It reads the raw conversation records (transcripts) that various AI programming tools have already written locally, and presents the scattered data in a single dashboard, solving the problem that built-in statistics tools miss sessions, omit entire days of data, and only count their own tool.
In terms of purpose, it helps users see the true scale of their AI usage: aggregating usage across all supported tools, and providing a trophy system unlocked based on usage. All processing is completed 100% offline, and data never leaves the local machine.
Main capabilities include:
- Cost estimation: estimates the dollar value of AI compute used based on public API prices, a metric many people are willing to screenshot and share.
- Tokens processed: measures the actual work the model truly completes, avoiding inflation caused by repeated cache reads, and provides a concrete reference such as "roughly equal to N times The Lord of the Rings."
- Average focus duration: the real active time spent with AI each day.
- AI personality profile: for example "Deep Diver, Night Owl," generated based on real usage patterns, with every conclusion showing the numbers behind it, without exaggeration or fabrication.
- Percentile ranking: based on a documented, fully offline-computed usage benchmark, gives a reference such as "in the top X% of AI developers," always accompanied by numerical evidence, without using fabricated rankings.
- Achievement system: three major categories—Volume, Intensity, and Mastery—with sixty tiered trophies in total, twenty per category, from First Steps to Billion-token clubs, from Night Owl to Whale. Token-based trophies count "processed (real work) tokens," not inflated billing figures, and are displayed on tiered badges by actual date earned. The three rarest achievements will appear at the top of the share card.
- Activity calendar: displays daily usage in a real calendar format.
- One-click share card: generates an image with branded statistics and personality profile, making it easy to share.
Supported tools and models: supports Claude Code, Codex, Kimi, Qwen CLI, HashCortx, HashCerebrum. Model names and public list prices can recognize Claude, GPT/Codex, Kimi, Qwen, GLM, MiniMax, Gemini and other series, so even non-Anthropic models running in Claude Code can be correctly named and priced, rather than being lumped together. Adding a new tool requires only one Rust file; the architecture uses pluggable source adapters, and PRs are welcome.
How to use: You can download the installer for your operating system from the Releases page, or build from source. Because it is a free and open-source project and is unsigned, macOS and Windows may show security warnings. On macOS, open the .dmg and drag it into Applications. If a verification warning appears on first launch, right-click (or Control-click) the app → Open → Open, or go to "System Settings → Privacy & Security" and choose "Open Anyway"; if it says "damaged and can't be opened" (common on Apple Silicon due to the quarantine attribute), run xattr -dr com.apple.quarantine /Applications/HashMeterAi.app in Terminal and then reopen it. On Windows, run the .msi / .exe installer, and when SmartScreen appears click "More info → Run anyway"; if the browser blocks the download, choose to keep it. Linux provides .AppImage or .deb: for AppImage, first run chmod +x HashMeterAi_.AppImage and then run it; if it won't start, install FUSE (on Debian/Ubuntu, sudo apt install libfuse2). For .deb, install with sudo dpkg -i HashMeterAi_.deb, then run sudo apt-get install -f to fill in dependencies. Building from source requires Rust (stable) and Node 18+, and skips all the above system warnings.
Technical architecture: Uses Tauri v2 + Rust + native JavaScript, does not use Electron, and does not use a bundler. Produces a small native binary (target under 15 MB). All parsing and aggregation are done in Rust (parallel and incremental, with instant refresh). The interface is pure HTML, CSS, JS.
Hash ecosystem: HashMeterAi is one of three local-first, privacy-first applications by the same developer, with no cloud and no telemetry; data stayson the local machine. HashCortX and HashCerebrum write local token usage logs, and after HashMeterAi reads them, it can accurately measure usage across the entire ecosystem in one place.
Limitations and notes: Cost figures are estimated compute value based on public API list prices, not a bill; under subscription plans, you are not paying per token. Token counts are read from each tool's local files and may differ from the provider's official dashboard. "Top X%" is a percentile calculated locally based on a static, modeled 2025–26 usage baseline, labeled as a modeled baseline, and is not a real-time leaderboard. All product names, logos, and brands belong to their respective owners and are used only for identification. HashMeterAi is not affiliated with, endorsed by, or sponsored by Anthropic, OpenAI, Google, Alibaba, Moonshot AI, Z.ai, MiniMax, or other mentioned tools. The software is provided "as is" without any warranty. The license is Apache-2.0, copyright Seif Hashish (Hash-7777) 2026. All rights reserved.
