Do I need an API key to check Claude Code usage?
No. Claude Code already writes token counts to local session files on your Mac, so a tool can read them directly. Tokens 4 Breakfast auto-detects these files with no API key and no account.
How to check Claude Code usage on a Mac
Claude Code records every session to your Mac. That means you can see exactly how many tokens you are using — by project and model — without an API key or a cloud dashboard. Here are the three ways to do it, from a quick terminal check to an always-on menu bar view.
Quick answer
Claude Code writes each session to local JSONL files in ~/.claude/projects/, including exact token counts. To check your usage you can read those files with a CLI like ccusage (npx ccusage) for a one-off table, or use a menu bar app like Tokens 4 Breakfast, which auto-detects the files and shows tokens by project and model live — no API key and nothing leaves your Mac.
Step by step
The full walkthrough
Each step stands on its own. Skip to the one that matches where you are.
Find your local session files
Claude Code stores a JSONL log for every session under ~/.claude/projects/, organized by project folder. Each line records the model used and exact input, output, and cache token counts. Nothing here needs an API key or a login — it is already on your Mac.
Decide between a terminal snapshot and an always-on view
If you just want a quick number, a terminal tool that parses those logs is enough. If you want to notice usage while you work — before you hit a limit — a menu bar app that keeps the number visible is the better fit. Most heavy users end up wanting the second.
Quick check: run ccusage in the terminal
ccusage is a popular open-source CLI. Running npx ccusage reads your local Claude Code logs and prints a table of tokens and estimated cost by day and model. It is great for a one-off audit, but it only shows what you ask for, when you ask for it.
Live view: install Tokens 4 Breakfast
Download Tokens 4 Breakfast, open it, and it auto-detects ~/.claude/projects/ with no setup or API key. Your Claude Code usage appears in Usage Intelligence — tokens broken down by project, model, and cache — and the live number sits in the menu bar so it is always one glance away.
Read the breakdown that actually explains your burn
Look at input vs output vs cache-read tokens, and which project and model are consuming the most. Opus-heavy sessions and large contexts dominate weekly usage; cache reads are cheap. This breakdown is what tells you where to cut, not just that you are close to a limit.
Pro tips
- Cache-read tokens are far cheaper than fresh input tokens — a project that looks expensive may mostly be cache reads.
- If a single project dominates your usage, it is usually a large or repeatedly re-read context, not the number of prompts.
- Turn on rate-limit alerts so you get a heads-up before the 5-hour or weekly window caps you mid-task.
FAQ
Common questions
Short, direct answers to the things people ask most about this.
Where does Claude Code store usage data on a Mac?
In ~/.claude/projects/, as JSONL files organized by project. Each line includes the model and exact token counts for that turn.
Is checking usage this way private?
Yes. Reading local session files happens entirely on your machine. Tokens 4 Breakfast is local-first by design — no cloud, no telemetry, and a built-in audit log of any network activity.