Comparison
Both live in your menu bar and watch your AI coding tools — but they answer different questions. CodexBar answers “how much do I have left right now?” — rate-limit windows, reset countdowns, and credit balances across 50+ providers. TokenMaxing answers “what did we actually use, and what did it cost?” — persistent usage and cost history, attributed per repo, per developer, and per account, with teams and OpenTelemetry forwarding.
Full disclosure: TokenMaxing exists because we needed several Claude and Codex accounts tracked separately on one machine — and nothing did it.
Pick CodexBar if
You want a free, open-source menu-bar gauge of what’s left — session and weekly limit windows, reset countdowns, credit balances, and provider status across dozens of AI tools on your Mac.
Pick TokenMaxing if
You want usage tracked over time and attributed per repo, per developer, and per account — even multiple Claude or Codex accounts on one machine — compared on a leaderboard, rolled up across a team, or streamed as cost-enriched OpenTelemetry into your own observability stack.
| Feature | CodexBar | TokenMaxing |
|---|---|---|
| Price | Free, open-source (MIT) | Free to start |
| Runs | macOS menu bar (14+) + CLI | Local menu-bar app + optional cloud |
| Setup | brew install --cask codexbar | Install the desktop app |
| Core job | Rate limits & reset countdowns | Usage, cost & attribution history |
| Providers covered | 50+ (Codex, Claude, Cursor, Gemini…) | Claude Code & Codex |
| Session / weekly limit windows | Yes | No |
| Credit balances & reset credits | Yes | No |
| Provider status & incident badges | Yes | No |
| Cost in USD | Local cost scans (Codex & Claude) | Yes |
| Multiple Claude / Codex accounts, one machine | One account per provider | Tracked separately per account |
| Per-repo attribution | No | Yes |
| Per-developer / multi-user | No | Yes |
| Team & org rollups | No | Yes |
| Persistent history across machines | No | Yes |
| History survives local log cleanup | No | Yes |
| Leaderboard & badges | No | Yes |
| Shareable profile card | No | Yes |
| Cost-enriched OpenTelemetry forwarding | No | Yes |
| No account required | Yes | Yes |
| Raw logs stay 100% local | Yes | Yes |
| Cloud data | None | Usage metadata only, when enabled |
CodexBar is an independent open-source project by Peter Steinberger. Comparison reflects general capabilities and may change as both tools evolve.
Keep the history, not the files
Agent session stores exist on your Mac and can be rotated or cleaned as they age. Claude Code, for example, deletes session files older than 30 days by default.
Once TokenMaxing syncs the normalized usage metadata, your historical cost and attribution remain available in the cloud—even after local source files are gone. Raw logs, prompts, transcripts, and source code never leave the device.
CodexBar is a free, open-source macOS menu-bar app that shows how much of your AI coding limits remain — session and weekly windows, reset countdowns, and credit balances — across 50+ providers like Codex, Claude, Cursor, and Gemini. It adds provider status badges, home-screen widgets, and a bundled CLI. It’s a live gauge for one Mac: no account, and your data stays local.
TokenMaxing tracks Claude Code and Codex usage and cost, attributes every token to a repository and a developer, and keeps a persistent history — including when you run multiple Claude or Codex accounts on one machine, each tracked as its own identity. Its local menu-bar view requires no account, and raw logs remain on the device. Optional cloud features sync only normalized usage metadata for profiles, private team views, and cost-enriched OpenTelemetry forwarding.
They answer different questions. CodexBar tells you how much quota you have left right now — session and weekly windows, reset countdowns, and credit balances in your Mac menu bar. TokenMaxing records what you actually used over time and attributes it per repository, per developer, and per account, with team rollups and OpenTelemetry forwarding. Plenty of developers run both: CodexBar to pace the day, TokenMaxing for the history and the receipts.
Yes. CodexBar is free and open-source (MIT), built by Peter Steinberger. Install it via Homebrew or grab the release from GitHub.
No — that’s CodexBar’s home turf. TokenMaxing focuses on what you used and what it cost, not how much quota remains in the current window.
Yes — this is a core TokenMaxing feature. Point the collector at each account’s config directory (say, a personal Claude and a work Claude) and every account is tracked as its own identity, with its own history and attribution. CodexBar shows one signed-in account per provider.
No. CodexBar’s local cost scans estimate total spend for Codex and Claude on that Mac, but nothing is attributed to a repository or developer, and there are no team rollups.
Yes. TokenMaxing is OpenTelemetry-native and can forward OTLP into any OTLP-compatible observability stack, with estimated USD cost injected per token. CodexBar is a menu-bar gauge and doesn’t emit telemetry.
No. The local menu-bar view works without an account, and raw logs, prompts, transcripts, and source code remain on your machine. If you sign in and enable cloud features, only normalized usage metadata is synced.
TokenMaxing keeps normalized usage metadata that was already synced, so historical usage and cost trends remain available even after local source files are rotated or deleted. For example, Claude Code deletes session files older than 30 days by default. The raw files themselves are never uploaded.
Only if you want it to be. Raw logs always remain local, synced usage metadata is private by default, and the global leaderboard is opt-in. Team dashboards are private by default.
Keep exploring
Keep the menu-bar gauge — then let TokenMaxing keep the history, the attribution, and the cost math. Free to start.