Home / Claude Code & Codex cost
How to see what Claude Code and Codex cost and how many tokens they use
In Claude Code, type /usage (/cost and /stats are aliases): it shows the session’s estimated cost, lines changed and tokens per model, and on a Pro or Max plan your usage limits. In Codex, type /status for this session’s token usage, context left and your 5-hour and weekly limits, or /usage for daily and weekly token activity. For totals across days, projects and both agents, read the logs they keep on your computer: npx ccusage@latest prints token and cost reports, and DANGO keeps a running count in your menu bar and also estimates what the work was worth.
Claude Code
| Command | What it shows |
|---|---|
/usage (or /cost, /stats) | The current session: estimated cost at API list prices, duration, lines changed, input / output / cache tokens per model. On Pro, Max, Team and Enterprise plans: how much of your plan limits you have used, and what used them. |
/context | How full the context window is, as a coloured grid. |
/status | Version, model, account and connection; no cost or tokens. |
Anthropic’s docs note that the session cost figure is meant for API users; on a Pro or Max subscription it isn’t what you are billed. API users can also see spending in the Claude Console and export claude_code.cost.usage and claude_code.token.usage through OpenTelemetry (CLAUDE_CODE_ENABLE_TELEMETRY=1). Sources: Manage costs, Monitoring.
Where the logs are: every session is a JSONL transcript under ~/.claude/projects/<project>/. Each assistant reply carries its model and a usage block (input, cache writes, cache reads, output). Note that one reply can be written as several lines with the same message ID; count it once.
Codex
| Command | What it shows |
|---|---|
/status | Model, approval mode, context window left, this session’s token usage, your 5-hour and weekly limits and credits. |
/usage | Your ChatGPT token activity by day, week or in total. |
/statusline | Puts model, context, limits and token counters in the footer. |
Where the logs are: ~/.codex/sessions/YYYY/MM/DD/rollout-….jsonl (older ones move to ~/.codex/archived_sessions/). Each rollout records token counts with a running session total and the last response’s usage (input, cached input, output, reasoning). Sources: Codex commands, openai/codex.
Across days, projects and agents
The built-in commands answer “this session” and “my plan”. To see last month, or Claude Code and Codex side by side, you read the local logs:
- ccusage (MIT, command line):
npx ccusage@latest dailyprints tokens and cost by day, month or session, for Claude Code, Codex and other agents;ccusage codex dailyfor Codex alone. - DANGO (free Mac app, Windows preview): reads the same logs in the background and keeps today, this week, this month and all-time in your menu bar, per agent. Next to the model cost it shows what the work would cost to hire out, so you see leverage, not just spend. How that is estimated →
- Menu bar limit meters such as CodexBar or ClaudeBar show how close you are to your plan’s limits. Compare them →
What “cost” means on a subscription
On Claude Pro or Max, or ChatGPT Plus or Pro, you pay a flat price, not per token. Tools that read the logs, DANGO included, price the tokens at the official API list prices: what the same usage would cost through the API. That is useful for comparing days, agents and models, but it isn’t your bill. Your bill is the plan price.
Questions
What is the Claude Code command to see cost?
/usage. /cost and /stats are aliases for it. It shows the current session’s estimated cost and tokens per model; on Pro and Max plans it also shows your usage limits.
How do I see token usage in Codex?
Type /status for the current session’s token usage, context left and your 5-hour and weekly limits, or /usage for daily and weekly activity.
How do I see Claude Code usage for the whole month?
The built-in command covers the current session and your plan limits. For a month, read the local logs in ~/.claude/projects with npx ccusage@latest monthly, or let DANGO keep the totals in your menu bar.
Is the cost shown accurate if I’m on Claude Max?
It is the API list price of your tokens, not what you pay. On Max you pay the plan price, so the figure is a way to compare usage, not your bill.
DANGO is a free Mac menu bar app that tallies how much work Claude Code, Codex and other AI tools did for you, what it cost and what it's worth.
Or install from Terminal: curl -fsSL https://dangoai.org/install.sh | sh
Free · macOS 13+ · Windows 10+ (preview) · Runs on your computer · GitHub