Track Claude Usage from Your Mac Menu Bar

What is Claude?

Claude is Anthropic's family of AI assistants, available through the claude.ai web and mobile interface and the Anthropic API. The model lineup includes Claude Haiku (fast and cost-efficient), Claude Sonnet (the best all-around for most tasks), and Claude Opus (highest capability for complex reasoning). Claude has become the default AI assistant for many software developers, particularly those using Claude Code — an agentic coding tool that runs directly in the terminal and interacts with files, runs shell commands, and executes multi-step tasks autonomously.

Unlike some competitors, Anthropic publishes relatively few details about exact rate limit thresholds, which makes tracking your own usage especially important. The limits vary by product (claude.ai chat vs. API vs. Claude Code), by subscription tier, and by the specific model you're using.

Claude Rate Limits: A Full Breakdown

Claude's rate limits work completely differently depending on which product you're using. Conflating them is the most common reason developers hit unexpected walls.

claude.ai (chat interface)

The web interface uses a message-based limit inside a rolling 5-hour window:

• Free: Limited Sonnet access, primarily Haiku. Throttles after roughly 20–30 Sonnet messages per day.
• Pro ($20/mo): Approximately 45 messages per 5-hour window on Sonnet 3.5 / 3.7. Unlimited Haiku messages.
• Max 5× ($100/mo): Approximately 225 Sonnet messages per 5-hour window.
• Max 20× ($200/mo): Approximately 900 Sonnet messages per 5-hour window.

The 5-hour window is rolling, not clock-aligned. If you send your first message at 2:17 PM, the window resets at 7:17 PM — not at 7:00 PM. Claude doesn't show you the exact reset time; you're left guessing.

Claude API

API limits are token-based (TPM — tokens per minute) and request-based (RPM — requests per minute). They're set per organization and scale with your account tier (Tier 1 through 4). New organizations start with conservative Tier 1 limits and graduate to higher tiers after reaching cumulative spend thresholds.

Claude Code

Claude Code calls the Anthropic API directly — not claude.ai. There's no 5-hour message window. Instead, you're drawing from your organization's API token budget, billed per token at the end of the month. A single agentic Claude Code session that reads files, runs commands, and iterates on edits can consume tens of thousands of tokens in minutes.

Why Tracking Claude Usage Matters

Claude's 5-hour window resets silently. There's no push notification, no email, no visible countdown. The claude.ai interface shows a rough "you're nearing your limit" message, but it doesn't tell you how many messages you have left, how many minutes until your window resets, or whether you can fit in one more task before the cutoff.

For developers using both claude.ai and Claude Code in the same session, the situation is worse. The two usage pools are completely independent — but without a unified view, it's easy to think you have capacity when you're actually approaching the API limit, or vice versa.

Getting rate-limited mid-task is one of the most disruptive things that can happen during a coding session. You lose the thread of what you were doing, context resets, and you either wait for the window to reset or switch to an inferior fallback model. Knowing you're at 80% capacity five minutes in advance means you can wrap up cleanly instead of getting cut off.

How AIUsageBar Tracks Claude

AIUsageBar shows your Claude usage in a persistent menu bar popup that's always one click away. For claude.ai Pro and Max users, it displays your current session message count, the percentage used in the current 5-hour window, and the exact number of minutes until your window resets — not an approximation, not a vague warning, but a real countdown.

For Claude Code users on the Pro plan, AIUsageBar also tracks your Anthropic API token spend for the day and the month, so you can see how fast you're burning through your budget in real time rather than discovering the damage at billing time.

Frequently Asked Questions