Ollama Cloud usage limits on your Mac — UsageNow
Keep your Ollama Cloud session and weekly limits in the macOS menu bar with UsageNow, using a Cloud API key kept in your Keychain and sent only to ollama.com. Experimental.
See usage limits, reset times, credits, and the last 30 days of activity for Codex, Claude Code, Antigravity, Gemini CLI, Kiro, Warp, OpenCode, and Qoder — plus DeepSeek, Kimi, OpenRouter, and Ollama Cloud — in one clear view in the Mac menu bar.
What UsageNow shows
UsageNow is a native macOS menu bar app. Click its symbol and a popover shows every provider you track: each limit window, how much of it is left, and when it resets — alongside the tokens and requests recorded on your Mac today, by model, what those tokens would cost at published API prices, and a chart of the last 30 days. It works with Codex, Claude Code, Antigravity, Gemini CLI, Kiro, Warp, OpenCode, Qoder, DeepSeek, Kimi, OpenRouter, and Ollama Cloud, and Cursor, GitHub Copilot, and Qwen are coming soon. Limits appear exactly as a provider reports them: a 5-hour window, a weekly one, or both. When a value isn’t reported, UsageNow says so instead of inventing it. Features.
Ollama Cloud
Ollama Cloud’s session and weekly limits come from ollama.com with a Cloud API key you add, kept in the Keychain and sent to Ollama only. Ollama doesn’t document this, so the provider is marked experimental; if the answer changes, UsageNow shows the limits as unavailable rather than guessing. No reset time is reported, so none is shown. Ollama Cloud in UsageNow · Ollama Cloud documentation.
Codex
Codex usage limits, reset times, and your plan come from the official Codex app-server, which UsageNow runs locally, so it never reads or copies Codex credentials. Token and request counts come from Codex session files on your Mac. If the app-server can’t be reached, the last recorded limits are shown and marked as stale. Codex in UsageNow · Codex documentation.
Claude Code
Claude Code activity — tokens, requests, most recent model — and your plan are read from local session data. Usage limits are experimental and off by default: when you turn them on, UsageNow reads the Claude Code sign-in already in your keychain, keeps the token in memory only, and asks Anthropic for your usage. It never refreshes, modifies, or stores your credentials. Claude Code in UsageNow · Claude Code documentation.
Gemini CLI
Gemini CLI tokens, requests, and the models used today come from the session recordings Gemini CLI writes on your Mac. Only identifiers, timestamps, model names, and token counts are read; credential files are never opened. Gemini CLI doesn’t record usage limits locally, so UsageNow shows none rather than guessing. Gemini CLI in UsageNow · Gemini and Antigravity · Gemini CLI documentation.
Antigravity
Google Antigravity’s 5-hour and weekly limits — one for each group of models — and your plan come from the Antigravity app itself: while it runs, UsageNow asks the app’s own server on 127.0.0.1 for the quota summary its usage panel shows. No credentials are read or sent. When the app is closed, the last limits stay for a day, marked as stale. Antigravity in UsageNow · Antigravity documentation.
Kiro
Kiro’s monthly credit limit, its reset date, and your plan come from the usage answer Kiro itself writes to its log while it runs, so UsageNow makes no request to Kiro and reads no Kiro credential. Today’s credits and prompts come from Kiro’s session files. A free trial that has ended is ignored, so a plan with every credit left is never shown as used up. Kiro in UsageNow · Kiro documentation.
Warp
Warp’s agent requests, credits, and tokens by model for today are read from Warp’s own database on your Mac, opened read-only; the columns holding prompts and replies are never selected. Warp counts credits per conversation, so a conversation that began earlier is left out and the day is marked partial rather than overstated. Warp keeps no current credit limit on the Mac, so none is shown. Warp in UsageNow · Warp documentation.
OpenCode
OpenCode’s tokens, requests, and models, and what they would cost at API prices, come from OpenCode’s own SQLite database, opened read-only. SQLite hands over only each answer’s time, model, and token counts, so reply text stored beside them never leaves the database, and prompt tables aren’t read. OpenCode reports no limit of its own — limits belong to the provider you run it with. OpenCode in UsageNow · OpenCode documentation.
Qoder
The credits Qoder spent — today and over the last 30 days — and the answers they paid for come from the transcripts Qoder’s agent writes on your Mac, in the Qoder app or qodercli. Only each answer’s time, request identifier, and credits are decoded; prompts and replies are not, and Qoder’s sign-in is never opened. Qoder keeps your plan’s remaining credits only in its app, so no limit is shown. Qoder in UsageNow · Qoder documentation.
DeepSeek, Kimi, and OpenRouter
These pay-as-you-go accounts publish an official API for what’s left, so UsageNow reads them with an API key you add: DeepSeek’s balance in your account’s currency, Kimi’s available balance, and OpenRouter’s spending today and this month with the key’s limit. The key is kept in the Keychain on your Mac, sent over HTTPS only to that provider’s own host, used only for read requests, and deleted when you turn the provider off. Pay-as-you-go accounts in UsageNow · API key providers documentation.
Coming soon
Cursor, GitHub Copilot, and Qwen are listed in UsageNow as coming soon, with no integration yet. Each is added only when reliable, privacy-respecting usage data exists. Coming soon.
Features
- Menu bar: icon only, the most critical usage, or one provider you choose.
- Desktop widgets, small and medium, rendered from a sanitized snapshot the app publishes.
- Providers: Codex, Claude Code, Antigravity, Gemini CLI, Kiro, Warp, OpenCode, and Qoder read on your Mac, and DeepSeek, Kimi, OpenRouter, and Ollama Cloud read with an API key you add. Each can be turned off and reordered; Cursor, GitHub Copilot, and Qwen are coming soon.
- Credits: for tools that meter in credits — Kiro, Warp, and Qoder — instead of tokens.
- The last 30 days: a bar per day under each provider, with the month’s tokens or credits, estimated cost, and top model.
- Remaining or used: every percentage — popover, menu bar, and widget — reads the way you choose.
- Report a Problem: a diagnostic report you see in full, sent as a GitHub issue or email only by you.
- Balances: what’s left on a pay-as-you-go API account, in the currency the service reports.
- Activity by model: tokens and requests for each model used today, most active first.
- Team plans: your plan and seat, and only your own usage.
- Cost estimate: what today’s tokens would cost at published API prices — a sense of scale, not a bill.
- Updates: UsageNow installs its own signed updates, and a check sends nothing about you.
- Local-first: your usage data stays on your Mac, and there is no UsageNow account or server.
- Privacy: only timestamps, identifiers, model names, and token counts are read — never prompts, responses, or code. Sharing anonymous usage analytics is opt-in and off by default. Privacy documentation.
Getting UsageNow
UsageNow is free and open source under the MIT License. Version 0.6.0 runs natively on Apple silicon and Intel Macs with macOS 15 or later. Download the signed, notarized disk image from usagenow.com or GitHub Releases, or build it from source — see the installation guide. Cursor, GitHub Copilot, and Qwen and a command-line tool are coming. What’s new in 0.6.0.
More
Provider