beehiiv
You run npx ccusage and it tells you yesterday came to, say, $41 at API prices. On a Max plan that is interesting and not much use. You did not pay $41. You paid for the plan.
What you actually want to know is whether you will hit the limit before the refactor is done.
ccusage is a good tool. It is free, open source and fast, and it reads your local logs without uploading anything. It answers a cost question very well. On a plan, the question is the limit.
Which number a plan user needs
Anthropic’s own Claude Code docs say it plainly. Pro and Max subscribers have usage included in their subscription, so a session’s cost figure is not relevant for billing. What stops you is the five-hour window and the weekly limit.
The closest ccusage gets is its blocks report. It groups Claude Code usage into five-hour billing windows and shows the tokens and cost in each. It can warn you against a token limit, but that limit is a number you set, or the biggest block you have had so far. It does not come from Anthropic.
The AI usage widget reads what Claude Code and Codex last heard from their own servers. The percentage left of the current five-hour window and of the week, and when each resets. It is the same plan limit /usage shows in Claude Code, and the widget says how long ago each app checked, so a number is never older than it looks.
Always on screen
ccusage answers when you run it. Its live monitor was removed in version 18, and the statusline it offers sits inside Claude Code’s status bar.
The board answers when you glance at it. Two rings for Claude Code, two for Codex, green at half or more, amber from a fifth, red under that, each with its reset time. “Resets in 1h 42m.” “Resets Mon 9:00 AM.” When Claude is at 15% and Codex is at 80%, you know where the next task goes without typing anything.
The sessions waiting on you
The same runner sees every Claude Code and Codex session on your computer, in a terminal, VS Code or an app that runs them. Agent activity shows each one’s title and project, and whether it is working, done or waiting for you to answer a question, give an OK or approve a plan.
That last state is the expensive one. A session that waits half an hour for a yes is half an hour of work that did not happen. On the board, Needs you lists it.
What it costs a solo developer
ccusage is free. Context Board is $6 a month after a 3-day trial, and that one plan has the runner, the AI usage widget and every connection on the board. Set it up once with one line from Connect, and it starts with your computer from then on.
Against what you already pay for Claude or ChatGPT, $6 is small. One five-hour window you did not burn by surprise is worth more than that.
Where ccusage still fits
ccusage is built for the cost question, and it is excellent at it. Daily, weekly, monthly and session reports of tokens and estimated dollars, per-model breakdowns, JSON for your own scripts, and the same view across Claude Code, Codex, Gemini CLI, OpenCode, Amp, GitHub Copilot CLI and more. If you pay per token through the API, or you want to know which project burned the most, that is the report you want, and it is free under the MIT license.
That is an accounting job. Context Board shows the plan limit and the sessions, and it does not estimate dollars. Run ccusage when you want the bill in tokens. Keep the board open for the limit.
Try it for a normal day
Link your computer with the one line from Connect, then start from the Agents template, or add the AI usage and Needs you widgets from the gallery. Work the way you usually do. Watch the rings instead of waiting for the wall.
Start your Context Board, free for 3 days. The Claude Code usage and Codex usage pages show the widget in detail.