Updated · November 17, 2026

Claude Code & Codex Usage Limit Trackers: CodexBar vs ccusage vs Built-in Gauges

Subscription coding agents meter work across two separate dimensions: rolling 5-hour and weekly quota percentages enforced on provider servers, and raw tokens plus prompt-cache hits recorded locally in session logs. CodexBar reads live provider quota windows across 15+ tools; ccusage audits offline JSONL token spend and 5-hour billing blocks; native /usage, /status, and statusline payloads expose provider-reported numbers with zero third-party credential risk.

Direct answer

Match the tracker to the metric you actually need

Pick CodexBar when...

You run Claude Code, Codex CLI, Cursor, Gemini, or Antigravity on macOS or Linux and want live 5-hour session %, weekly cap %, and exact reset countdowns in your menu bar or CLI.

Pick ccusage when...

You want cross-platform token totals, cache-read efficiency, per-project breakdowns, and API-rate cost audits parsed strictly offline from ~/.claude and ~/.codex JSONL files.

Pick native /usage & statusline when...

Corporate policy prohibits third-party tools from reading Keychain OAuth tokens or browser session cookies, and you still want accurate 5h/7d rate-limit telemetry inside your shell.

CodexBar

steipete/CodexBar · macOS 14+ & Linux CLI

Telemetry source
OAuth / local browser cookies & CLI status
Primary output
Live 5-hour session & weekly % quota + reset timers

ccusage

ryoppippi/ccusage · npx ccusage@latest

Telemetry source
Local ~/.claude & ~/.codex JSONL files
Primary output
Token totals, model breakdown & dollar-equivalent cost

Side-by-side tracker matrix

CriterionCodexBarccusage (CLI / @ccusage/codex)Built-in (/usage, statusline, /status)
Primary question answeredHow much of my Claude Pro/Max or ChatGPT Plus/Pro 5-hour and weekly rate-limit cap remains, and when does it reset?How many input, output, cached, and reasoning tokens did I burn today/this month across projects, and what would that cost at API rates?Quick in-session check of my active terminal quota bar or custom tmux/shell statusline.
Data sourceReads local OAuth tokens, desktop/browser session cookies, or CLI RPC/status probes across 15+ providers (Codex, Claude Code, Cursor, Gemini, Copilot, Antigravity, OpenRouter).Offline parser scanning local ~/.claude/projects/*.jsonl and ~/.codex/sessions/*.jsonl transcripts; zero provider login or network cookies needed.Claude Code /usage command + statusline rate_limits JSON fields; Codex CLI /status summary.
Shows official 5h / weekly %Yes — surfaces exact percentage used and countdown reset timers in the menu bar and codexbar usage CLI JSON.No — local JSONL files record token counts per turn, not Anthropic or OpenAI subscription rate-limit buckets.Yes — shows live session and weekly allocation directly inside the active agent CLI.
Supported platformsmacOS 14+ menu bar app (Homebrew Cask) + Linux codexbar CLI. Windows users typically use community forks such as Win-CodexBar.Cross-platform Node.js/Bun CLI (npx ccusage@latest, ccusage blocks, @ccusage/codex) across macOS, Linux, and Windows.Every OS where Claude Code or OpenAI Codex CLI runs.
Privacy & credential surfaceRequires granting permission to read local Keychain / browser cookies or spawning provider CLI probes; inspect permissions before enabling auto-cookie sync.Lowest permission surface: strictly reads local JSONL session logs (fetches public model pricing tables only when pricing is outdated).Zero third-party tooling outside Anthropic or OpenAI CLI.

Other open-source quota trackers worth knowing

lpm (lpm.cx)

Lightweight macOS bar focusing on real-time Codex and Claude burn-rate pacing so long tasks do not hit unexpected mid-session throttles.

ReserveGauge

Native Swift menu bar meter dedicated to Claude & Codex rolling quota reserves and reset clocks with minimal background overhead.

Claude-Code-Usage-Monitor

Python terminal dashboard featuring P90 session token velocity estimates and live countdown alerts before 5-hour rolling limits cap out.

Three real-world routing scenarios

Preventing mid-sprint lockouts during parallel worktree runs

Use CodexBar (or built-in statusline rate_limits JSON)

Token totals alone cannot warn you when a 5-hour rolling subscription window is at 88% because heavy reasoning models drain quota differently per plan tier. Keep a live percentage gauge visible before launching multi-subagent refactors.

Auditing spend & prompt-cache hit rates across API vs subscription

Use ccusage (npx ccusage@latest daily / session / blocks)

When deciding whether a developer should stay on Claude Max / ChatGPT Pro or switch an automated pipeline to direct API keys, ccusage calculates exact API-equivalent token value and cache-read efficiency from local session logs.

Air-gapped or enterprise-restricted laptops

Use ccusage + built-in /usage and /status only

Environments that disallow third-party binaries inspecting Keychain OAuth tokens or browser session cookies can safely run ccusage offline against local JSONL transcripts alongside native CLI commands.

Quick installation commands

CodexBar (macOS Cask)

brew install --cask steipete/tap/codexbar

Installs the macOS menu bar utility and codexbar usage --json CLI. Linux CLI builds ship under GitHub Releases.

ccusage & @ccusage/codex

npx ccusage@latest daily && npx @ccusage/codex@latest daily

Reads local ~/.claude and ~/.codex JSONL files across macOS, Linux, and Windows with no browser cookies or OAuth extraction.

Primary sources · November 17, 2026

Repositories and official documentation

Frequently asked questions

Does ccusage show my remaining Claude Pro/Max or Codex 5-hour percentage cap?

No. ccusage parses local JSONL conversation logs (~/.claude and ~/.codex) to report token counts, cache hits, and API-equivalent dollar cost. Because subscription rate-limit percentages live on provider servers rather than local logs, use CodexBar or Claude Code’s built-in /usage and statusline rate_limits JSON to see remaining % quota.

Is there an official way to display Claude Code quota without third-party apps?

Yes. Inside Claude Code, run /usage for interactive quota status or read the rate_limits payload passed to custom statusline scripts (including 5-hour and 7-day percentage utilization). In OpenAI Codex CLI, run /status inside the terminal session.

Does CodexBar run on Windows?

The upstream steipete/CodexBar project targets macOS 14+ (menu bar app + CLI) and ships a Linux CLI (codexbar usage). Windows developers either track token usage cross-platform with ccusage / @ccusage/codex, check native /status and /usage commands, or use community Windows tray ports.

Are ReserveGauge, lpm, or Claude-Code-Usage-Monitor viable alternatives?

Yes for niche setups: lpm (lpm.cx) and ReserveGauge focus on lightweight macOS menu-bar burn-rate alerts, while Claude-Code-Usage-Monitor adds P90 burn-rate predictions in Python terminals. CodexBar leads on multi-provider coverage (15+ coding tools), while ccusage remains the standard open-source JSONL cost analyzer.

Related guides on AgentSkillsHub