Pinemeter · AI usage monitor + model broker
Know your limits. Route your next task.
Track AI usage in your Mac menu bar. Give Claude Code and Codex a model broker that chooses where to send tasks using your rules and remaining quota.
Free · macOS 14+ · Signed and notarized
Quota monitoring needs no command-line
setup.
Your accounts. One place to look.
See the limits you actually use.
Provider-reported usage, with work and personal accounts side by side.
Claude
Five-hour, weekly, and available model-specific limits for each connected account.
ChatGPT / Codex
The quota windows and model rows exposed by your provider, from your existing browser session.
Gemini API
API quota through your Google AI Studio key, when Google supplies numeric usage.
Meters turn red at 90%. Stale data turns grey, so an old reading never looks current.
Built for a glance
Keep working.
Keep limits in sight.
No browser tab to keep open. Your meters stay in the menu bar while you write, research, or code.
- Your accounts, in order.Up to 12 miniature meters mirror the popover.
- A warning before the limit.Set usage alerts and choose your refresh interval.
- A display that fits your Mac.Choose monochrome or a system-colour palette.
- Your status line, too.Read the local
~/.pinemeter/usage.jsonexport.
Make it yours
Your meters. Your colours.
Choose from six colour palettes, or keep your menu bar monochrome.
Three of six palettes · Real app renders · Sample data · Enlarged detail
Start with the stable app
Three steps to your first meter.
-
Install Pinemeter
Download the macOS app. Open the DMG and drag Pinemeter to Applications.
-
Sign into your providers
Open Claude or ChatGPT in your browser and sign into the accounts you want to monitor.
-
Connect your accounts
Click the Pinemeter icon, then choose Scan open browsers. Your usage appears in the menu bar.
For Gemini, add a Google AI Studio key in Settings. Safari import needs Full Disk Access; the app explains the permission.
They go to the provider they belong to. No Pinemeter account to create. Public source under the MIT License.
Read the source and security notesModel broker · Optional after install
Put that quota to work
for your coding agents.
Give Claude Code and Codex a local model broker. Choose which models should handle planning, coding, or review, using the quota Pinemeter already tracks.
- Rank models by task, starting from a preset.
- See backup choices when the preferred route is unavailable.
- Hand off new tasks to configured T3 Code instances.
How the broker works
-
Connect your accounts
Pinemeter tracks the quota available across your connected accounts.
-
Set up your agent
Enable the broker. Apply the setup prompt from the Instructions pane, then verify the setup.
-
Delegate a task
Your agent asks Pinemeter for a model based on your routing rules and available quota, then launches the task.
| Starting harness | Supported delegation |
|---|---|
| Claude Code | Native Claude agents, Codex through /codex-exec, or configured T3 Code
sessions. |
| Codex | Native Codex agents or configured T3 Code sessions. Direct native Claude delegation is unsupported. |
| Either, through T3 | Claude or Codex, subject to your configured accounts, availability, and routing rules. |
Pinemeter selects the route; your agent launches the task. Install the dispatch tools your
routes need, including /codex-exec or t3-dispatch. T3 is
optional.
Your agent must follow the setup instructions. Connecting the broker alone does not enforce routing or switch models in an existing conversation.
Pinemeter for macOS
Your limits. A glance away.
Download Pinemeter for macOSFree and open source · macOS 14+ · Signed and notarized
A few useful answers.
What is Pinemeter?
Pinemeter is a free, open-source macOS menu bar app from Pine IT. It shows provider-reported AI usage limits and reset times across your connected accounts. An optional broker helps coding agents choose models using available quota and your preferences.
Which AI usage limits can I monitor?
Claude subscription windows across multiple accounts, quota rows exposed by ChatGPT/Codex, and Gemini API quota when Google supplies numeric usage. Provider fields can change. Unavailable or stale data is distinguished from current readings.
Do I need a command-line tool or a Pinemeter account?
Neither is required for quota monitoring. Connect your existing provider accounts through the browser scan, or add a Google AI Studio key for Gemini. Your provider subscriptions and limits still apply. The optional broker needs the coding tools used by your configured routes.
Where do my credentials go?
Provider credentials use macOS Keychain and are sent to their respective providers. ChatGPT access tokens stay in memory. Routing decisions and audit history stay local. Pinemeter contacts providers for quota data, GitHub for releases, and its configured HTTPS preset service for broker presets. Safari browser import needs Full Disk Access; the app explains the permission.
What is the broker, and how do I set it up?
Pinemeter includes an optional local Model Context Protocol (MCP) broker for Claude Code and Codex, including handoffs to configured T3 Code instances. After installing Pinemeter, use the Instructions pane to connect your agent and verify its setup. T3 dispatch also needs the t3-dispatch integration. The broker selects routes for new tasks; it does not switch models inside an existing conversation, run the models, or give you extra quota.
Which Macs are supported?
Pinemeter requires macOS 14 Sonoma or later. The app is signed with Developer ID and notarized. Download the DMG, open it, and drag Pinemeter to Applications.
Pinemeter is a Pine IT project. Not affiliated with Anthropic, OpenAI, or Google.