CodexBar and CrewTower answer different questions

Where CodexBar is better

It is free and open source, under the MIT license, with a large community behind it: over 22,000 stars on GitHub.

It covers far more providers. CodexBar's site lists 106. CrewTower's usage strip covers two: Claude Code and Codex.

It keeps a spend history. CodexBar scans local Codex and Claude logs for cost, with history windows of rolling days, month to date or all time. CrewTower keeps no token or cost history.

It runs in more places. macOS 14 or later, plus a Linux desktop app and a command line. CrewTower needs macOS 15.

Widgets and remote hosts. CodexBar ships WidgetKit widgets, and its Agent sessions list can include macOS or Linux hosts reached over SSH. CrewTower has neither: it watches the agents on the Mac it runs on.

It speaks your language. The app and site are localised into 23 languages. CrewTower is English only.

CrewTower vs CodexBar

Where CrewTower is different

It is built on the state CodexBar leaves out. CodexBar's own design notes for its sessions list name “permission-waiting state” and “exact tmux pane focus” among the things that are out of scope. Those two are most of what CrewTower does.

It answers, not only shows. When Claude Code, Codex, Cursor, Gemini CLI, Qwen Code or opencode stops to ask, the card shows the command and you allow or deny it there. The agent's hook is waiting on a local socket and the answer goes back down it, so the terminal never has to come forward and nothing is typed for you.

It tells you once, then a little more, then stops. A waiting session opens the notch briefly, nudges again at one, five and fifteen minutes, then goes quiet. A Focus mode silences it while the badge still shows who is waiting.

It lands on the exact place. Jump selects the tab in Terminal, iTerm2 and Ghostty and the pane in tmux and herdr, not just the app. How Jump finds the tab is written up terminal by terminal.

Desktop apps as well. Claude Code in the Claude app and Codex in the ChatGPT app get cards next to the terminal sessions, and when the ChatGPT app itself asks you something, such as Plan mode's “Implement plan?”, the notch announces that too.

Using both

They do not compete for anything. CodexBar sits in the menu bar with your usage and spending across every provider you use; CrewTower sits in the notch with the sessions and what they are waiting for. CrewTower's own usage strip is on by default and can be switched off in its settings if you would rather read the number in CodexBar alone.

Both read the Claude and Codex figures from the providers' own reporting, so expect them to agree; CrewTower makes no claim to a better usage number than CodexBar's.

Which one to pick

Only CodexBar if usage and spend are the whole question, if you use providers beyond Claude and Codex, or if you are on macOS 14 or Linux.

Both if you run several agents at once and the time you lose is the time one sits waiting for an answer in a tab you are not looking at.

A free notch app instead of CrewTower if you want the waiting sessions too but would rather not pay: CodeIsland and Open Island are free and open source and answer permission prompts from the notch. The comparison table has both, with the rows they win.

Questions

Is there a CodexBar alternative for Windows or Linux?

Not CrewTower: it is a Mac app and needs macOS 15 or later, on Apple silicon or Intel. CodexBar itself has a Linux desktop app and a command line, and its README links community ports for Windows.

Does it also work if I have several Claude sessions running at once?

Yes. Every running Claude Code session gets its own card, in terminals and in the Claude app, and the ones waiting on you sort to the front. One card can also hold several requests at once, from the session and its subagents, each answered on its own.

On the card I only see “working”. Can I see what it is doing right now, like the last tool call or a progress line?

Yes. Press Activity on the card: it opens the last stretch of that session's transcript inside the card, what the agent said and what it ran, in order, newest at the bottom. For the agents CrewTower follows live, the line under the title also names the current step, and running subagents are listed beneath it.

Sources

Read on 19 October 2026: codexbar.app for the price, providers and macOS; github.com/steipete/CodexBar for the license, the Linux app and command line, the Windows note, widgets, cost scans, languages and the release (0.78.0, 19 October 2026); docs/agent-sessions-design.md for the non-goals, quoted above; docs/sessions.md for the sessions list, SSH hosts and its default; docs/claude.md and docs/codex.md for where its figures come from; docs/cost-reporting-periods.md for the history windows.