cihanatak

BurnMeter — Data skill for Claude Code

Data community

Local-first usage & cost dashboard for Claude Code + OpenAI Codex — your burn rate, rate-limit proximity, and where every token goes.

How to install BurnMeter

This entry records only its repository, not the path inside it, so there is no exact command to give. Open cihanatak/BurnMeter and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What BurnMeter does

Local-first usage & cost dashboard for Claude Code + OpenAI Codex — your burn rate, rate-limit proximity, and where every token goes. No API keys, no cloud, data never leaves your machine.

Alternatives in Data

Data Contract — This document defines which files belong to the system (auto-updatable) and which belong to the user (never to 32.5k ★
Claude Code Usage Monitor — by Maciek-roboblog - A real-time terminal-based tool for monitoring Claude Code token usage 7k ★
Token Dashboard — See where Claude Code is burning tokens - turn raw JSONL transcripts into local cost analytics, hotspot views 669 ★

README

Burnmeter

**One dashboard for every AI coding agent — see exactly where your tokens go.**

A **local-first** dashboard that reads the session logs your AI coding agents already write to disk and turns them into a real-time view of token usage, cost, burn-rate, and — where the agent actually logs it (e.g. Codex) — **how close you are to your plan's rate limit**. No API keys. No cloud account. No phone-home. Your data never leaves your machine.

![The Burnmeter dashboard — burn-rate gauge, rate-limit hero, cache savings, per-project & per-model breakdowns, activity heatmap](docs/dashboard.png)

burnmeter.dev · pipx install git+https://github.com/cihanatak/BurnMeter · Claude Code + OpenAI Codex · AGPL-3.0

ℹ️ Screenshot uses synthetic demo data (`tests/generate_fixture.py`) — your real numbers never leave your machine.

Why this exists

Neither Anthropic nor OpenAI exposes a real-time, per-token usage API for **individual** Pro / Max / ChatGPT-subscription plans. The in-app usage page and the CLI's `/usage` command are point-in-time only — not historical, not programmatic. So if you're on a flat-fee subscription, **your local session logs are the only accurate source of truth** for what you're actually burning.

Burnmeter reads them — `~/.claude/projects/**/*.jsonl` and `~/.codex/sessions/**/*.jsonl` — and answers the one question that matters: **"Am I about to hit the wall?"**

Features

Binding-constraint hero — the single most-urgent number, front and center. Codex logs real rate-limit usage, so its hero is your rate-limit % (5-hour + weekly) — red ring at 90%+ means "about to get throttled." Claude Code exposes no account-level limit (Anthropic doesn't write one to disk), so its hero is burn-rate $ + live local usage — honestly labelled, never a guessed limit.
Real per-model pricing — Opus / Sonnet / Haiku and GPT-5.5 / 5.4 / Codex