HashMeterAi tracks real AI coding tool usage locally with no data leaving your machine
HashMeterAi is a local-first, fully offline usage meter that reads raw transcripts from AI coding tools like Claude Code, Codex, Kimi, and Qwen CLI to show unified token usage, estimated costs, and activity stats — all without sending any data to a server.
Score breakdown
HashMeterAi fills a gap left by per-tool built-in meters — which the project says skip sessions and only count themselves — by providing a single cross-tool view of real AI coding usage without requiring any data to leave the local machine.
- 01Reads raw local transcripts from Claude Code, Codex, Kimi, Qwen CLI, HashCortx, and HashCerebrum to unify usage data.
- 02Runs 100% offline — no server, no data leaves the user's machine.
- 03Shows estimated dollar value of AI compute at public API rates.
HashMeterAi, posted to Hacker News by Hash7777, is an open-source tool hosted at `Hash-7777/HashMeterAi` on GitHub under the Apache-2.0 license. It positions itself as a corrective to the built-in usage meters of individual AI coding tools, which the project claims skip sessions, miss whole days, and only count their own activity. Instead, HashMeterAi reads the raw local transcripts that tools like Claude Code, Codex, Kimi, Qwen CLI, HashCortx, and HashCerebrum already write to disk, then aggregates them into a single dashboard — entirely offline, with no server communication.
It also provides a percentile ranking computed against a documented offline benchmark modeled from public 2025–26 figures.
The dashboard surfaces several metrics: an estimated dollar value of AI compute consumed at public API rates, processed token counts (described as the "truest measure of work the model actually did," explicitly not inflated by cached re-reads), average focus time per day, and a generated "AI persona" such as "Deep Diver, Night Owl" backed by the underlying numbers. It also provides a percentile ranking computed against a documented offline benchmark modeled from public 2025–26 figures. A gamification layer adds 60 ranked trophies across three categories — Volume, Intensity, and Mastery (20 each) — ranging from "First Steps" to "Billion-token clubs," all based on processed rather than billed tokens. A daily activity calendar and a one-click shareable image card round out the feature set. The repository had 11 stars and 1 fork at the time of posting.
Key facts
- 01Reads raw local transcripts from Claude Code, Codex, Kimi, Qwen CLI, HashCortx, and HashCerebrum to unify usage data.
- 02Runs 100% offline — no server, no data leaves the user's machine.
- 03Shows estimated dollar value of AI compute at public API rates.
- 04Token counts use 'processed' tokens, explicitly excluding cached re-reads to avoid inflation.
- 05Generates an 'AI persona' (e.g., 'Deep Diver, Night Owl') derived from real usage patterns.
- 06Includes 60 ranked trophies across Volume, Intensity, and Mastery categories (20 each).
- 07Percentile ranking is computed against a documented offline benchmark modeled from public 2025–26 figures.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 16, 2026 · 23:11 UTC. How this works →