Best Codex Usage Trackers for macOS (2026)

Why a message count doesn't tell you where you stand

OpenAI publishes Codex allowances as ranges of messages per usage window rather than fixed counts, and explains why. Per the Codex pricing documentation: “The number of messages you can send depends on the model used, size and complexity of your tasks, and whether you run them locally or in the cloud.” The same page adds that small scripts may consume only a fraction of an allowance, while larger projects, long-running tasks, or extended sessions that require the agent to hold more context use significantly more.

One shared window is documented, with a longer one possible on top. From the plan rate tables: “The usage limits for local messages and cloud chats shared a five-hour window until it was suspended on 12 July 2026. Additional weekly limits may apply.” So local messages and cloud chats draw on the same five-hour window rather than two separate ones, and a weekly limit may sit above it.

Elsewhere on the same pricing page - as separate notes, not part of that footnote - OpenAI states that Enterprise and Edu users with flexible pricing have no fixed rate limits because usage scales with credits, and that Enterprise and Edu plans without flexible pricing have the same per-seat usage limits as Plus for most features.

Past the included allowance, the unit changes. OpenAI: “Credits translate token usage into a simpler unit for tracking and managing consumption. The credit cost varies by model, context, reasoning, and tools. After you reach your included limits, available credits let you continue working. Usage is calculated in credits per million input tokens, cached input tokens, and output tokens.” OpenAI publishes those credit rates per million tokens but no dollar-per-credit conversion in that source, so we do not convert credits into dollars anywhere on this page.

Speed settings change the burn rate too. OpenAI states that speed configurations increase credit consumption for all applicable models, so they also use included limits faster, and that image generations use included limits roughly 3-5x faster on average than similar turns without image generation, depending on image quality and size. The Speed documentation gives the multipliers per model.

So the practical question is not “how many prompts have I sent” but “how much of this window have I spent, and when does it reset.” Below are four places that answer some part of that, with what each is documented to show.

#1: SessionWatcher for Codex Our own product

SessionWatcher for Codex is a native macOS menu bar app for keeping Codex usage visible outside the terminal - Codex is one of the tools listed on our homepage. Our own materials describe the following:

• A persistent menu-bar reading of how much of the current window has been used, without opening a browser tab or running a command.
• A live countdown to when capacity returns.
• Coverage of the longer window alongside the five-hour one - OpenAI documents that “additional weekly limits may apply.”
• An estimated cost figure. Treat any locally computed figure as an estimate; your account's own billing view settles what you actually pay.
• Configurable macOS notifications at thresholds you choose, so the warning arrives before the interruption.
• Plan awareness for the ChatGPT plan tiers OpenAI lists on its Codex pricing page. We do not reproduce the per-plan message ranges here: OpenAI publishes them as ranges and revises them.

Setup: install the app and enter a license key. Requirements: macOS 14 or later; macOS only.

Pricing: $6.99 one-time for Codex tracking. Pro covers all supported tools and is sold as a one-time purchase - no subscription required - with $24/year as an alternative. Pricing and support are first-party claims too: confirm the current requirements, supported tools, and price in our pricing section before buying.

#2: The Codex usage dashboard (official)

OpenAI's answer to “where can I see my current usage limits” is the Codex usage dashboard at chatgpt.com/codex/settings/usage. From the pricing documentation: “You can find your current limits in the usage dashboard.” The same passage recommends checking it “every week or two to understand your pace and remaining capacity.” That is what OpenAI documents: where to find your current limits. It does not document a live or real-time reading, and we make no claim about how quickly the dashboard reflects a session either way.

Pricing: free; included with the plan.

Limitations: it is a page you open, so it is not visible while you work. Notifications, an always-on reading, a live reading, and a per-session breakdown are not documented on the page cited. If you use Codex with an API key instead of a ChatGPT plan, OpenAI documents that you are billed for the tokens Codex uses based on API pricing; it also documents that image generation uses API pricing rather than included ChatGPT limits. In either case your API billing view is the one to reconcile against.

#3: Codex CLI /status

The in-terminal option, documented in the same passage: “If you want to see your remaining limits during an active Codex CLI session, you can use /status.” OpenAI's Codex CLI documentation lists it among the slash commands as “show current session configuration.” Those two descriptions - your remaining limits, and the current session configuration - are the whole of what OpenAI documents for it.

Pricing: free; built into the Codex CLI.

Limitations: it is documented for use during an active Codex CLI session, so it is a command you run rather than a reading that stays in view, and it tells you nothing when the CLI is closed. Notifications and a cost figure are not documented for it. The individual fields it prints are not enumerated in the documentation cited, so we do not list them here - check your own output rather than assuming a layout.

#4: ccusage (open source)

ccusage is an MIT-licensed CLI that describes itself as: “Analyze coding (agent) CLI token usage and costs from local data.” Its README states that it “reads local usage data from coding agent CLIs and turns it into daily, weekly, monthly, and session reports,” and lists Codex among its supported sources with the command ccusage codex daily. Alongside Codex the README lists Claude Code, OpenCode, Amp, Droid, Codebuff, Hermes Agent, pi-agent, Goose, OpenClaw, Kilo, Kimi, Qwen, GitHub Copilot CLI, Gemini CLI, and Grok Build CLI.

Setup: no install required - the README documents running it via a package runner, for example npx ccusage@latest.

Pricing: free, MIT licensed.

Limitations: it reports on local data you have already generated, so it is a reporting tool rather than a live warning system, and it is a command you run in a terminal. The README documents its five-hour blocks view as a Claude Code billing-window feature; it does not document an equivalent for Codex windows. Support changes between releases, so re-read the README before relying on a specific source.

Comparison: what each source documents

Sources for this table, all read August 15, 2026: OpenAI's Codex pricing documentation and Codex CLI documentation; the ccusage README. SessionWatcher cells come from our own product materials - pricing and support included - and were not independently tested.

How to choose

If you check in occasionally, the two official options are enough and cost nothing: /status while you are in the CLI, and the usage dashboard for pacing. If you want reports across several agent CLIs from local data, ccusage is free and documents a Codex source.

The gap none of those documents filling is a reading that stays visible while you work and warns you at a threshold you set. That is the workflow SessionWatcher for Codex is built for, at $6.99 one-time - disclosed as our own product, macOS-only, and not independently tested here. Try the free options first and buy only if the always-on reading is the part you keep missing.

Frequently Asked Questions

What is the best Codex usage tracker for Mac?

It depends on the workflow. For an official reading, OpenAI documents two: the Codex usage dashboard, and /status during an active Codex CLI session. For reports across several agent CLIs from local data, ccusage documents a Codex source. For a macOS menu-bar reading outside the terminal, SessionWatcher is our own product at $6.99 one-time; its capabilities are first-party claims that were not independently tested for this comparison.

Can I track Codex usage for free?

Yes. OpenAI documents both free options: “You can find your current limits in the usage dashboard. If you want to see your remaining limits during an active Codex CLI session, you can use /status.” ccusage is a free MIT-licensed CLI that documents a Codex source and reads local usage data.

Does Codex meter messages or tokens?

Both units appear in OpenAI's Codex pricing documentation. Included allowances are published as message ranges per usage window (the weekly cap was suspended on 12 July 2026), and OpenAI states the number of messages you can send depends on the model used, the size and complexity of your tasks, and whether you run them locally or in the cloud. After the included limits, credits take over: “Usage is calculated in credits per million input tokens, cached input tokens, and output tokens,” and “The credit cost varies by model, context, reasoning, and tools.” That is why a message count alone does not predict when you will be cut off.

Do free Claude trackers like ccusage work with Codex?

ccusage does: its README lists Codex among its supported sources and documents a ccusage codex daily command. Claude-Code-Usage-Monitor does not; its README describes it as a companion for Claude Code. Re-check each README before relying on it, because open-source support changes between releases. See our how to check Codex usage guide for the full set of options.

How much does SessionWatcher for Codex cost?

$6.99 one-time for Codex tracking. Pro covers all supported tools and is sold as a one-time purchase - no subscription required - with $24/year as an alternative. These are first-party claims about our own product; the pricing section carries the current figures.