TokenSave
Claude Code Usage Analyzer
See how many tokens Claude Code, Codex and Cursor really used and what that would cost at API prices, by day, project and model, and whether your plan pays off.
🔒 Your files are read inside this browser tab. Nothing is sent to our server or anyone else. Close the tab and it's gone.
1. Open your usage logs
or drag the folder here
Where the logs are
%USERPROFILE%\.claude\projectsIn the folder picker, paste this into the address bar and press Enter.
~/.claude/projectsIn the picker press ⌘ + Shift + G, paste the path and press Enter. (⌘ + Shift + . shows hidden folders.)
~/.codex/sessionsCodex (CLI and app) logs: pick the sessions folder. On Windows paste %USERPROFILE%\.codex\sessions.
https://cursor.com/dashboardCursor doesn't keep token counts on your computer. Open cursor.com/dashboard → Usage → Export CSV and choose that file.
Chrome may ask “Upload files to this site?”. That's the browser's standard wording; the files stay on your computer.
2. Your usage
By tool
Cost per day
By project
By model
3. Plan or API?
Plan fees exclude VAT. Plans have 5-hour and weekly usage limits, so heavy use may hit a limit before the API cost reaches this level.
How it works
- Claude Code saves every conversation as a .jsonl file under ~/.claude/projects, one folder per project. Each reply has a usage block with input, output, cache-write and cache-read tokens and the model name.
- The same reply is often written on several lines (and copied again when you resume a session), so we count each reply once by its message and request ID.
- Each reply is priced at the official API rates for its model: base input, 5-minute and 1-hour cache writes, cache reads and output. Fast mode, US-only inference (1.1×), Haiku 5.5's long-prompt rate and web searches ($10 per 1,000) are included.
- By default Claude Code deletes logs older than 30 days. To keep more history, set "cleanupPeriodDays" in ~/.claude/settings.json.
- Codex writes a running token total into each session file under ~/.codex/sessions. We bill the increase between events at OpenAI's API prices, with cached input at the cached rate.
- Cursor no longer stores token counts on your computer, so we use the usage CSV from your Cursor dashboard and show the cost Cursor itself lists for each request.
API prices used (per 1M tokens)
| Model | Input | Output | 5m write | 1h write | Cache read |
|---|---|---|---|---|---|
| Claude Fable 5.1 / Mythos 5.1 | $10 | $50 | $12.50 | $20 | $0.25 |
| Claude Fable 5 / Mythos 5 | $10 | $50 | $12.50 | $20 | $1 |
| Claude Opus 5.5 | $4 | $20 | $5 | $8 | $0.20 |
| Claude Opus 5 · 4.8 · 4.7 · 4.6 · 4.5 | $5 | $25 | $6.25 | $10 | $0.50 |
| Claude Opus 4.1 / 4 | $15 | $75 | $18.75 | $30 | $1.50 |
| Claude Sonnet 5.5 | $2 | $10 | $2.50 | $4 | $0.10 |
| Claude Sonnet 5 | $2 | $10 | $2.50 | $4 | $0.20 |
| Claude Sonnet 4.6 · 4.5 · 4 | $3 | $15 | $3.75 | $6 | $0.30 |
| Claude Haiku 5.5 (≤100K) | $0.10 | $0.50 | $0.125 | $0.20 | $0.01 |
| Claude Haiku 4.5 | $1 | $5 | $1.25 | $2 | $0.10 |
| Claude Haiku 3.5 | $0.80 | $4 | $1 | $1.60 | $0.08 |
Codex: OpenAI API prices (per 1M tokens)
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6 Astra | $10 | $1 | $50 |
| GPT-6.1 Sol | $2 | $0.10 | $10 |
| GPT-6 Sol | $2 | $0.20 | $10 |
| GPT-6 Luna | $0.10 | $0.01 | $0.50 |
| GPT-5.6 Sol | $4 | $0.40 | $20 |
| GPT-5.6 Terra | $2 | $0.20 | $12 |
| GPT-5.6 Luna | $0.20 | $0.02 | $1.20 |
| GPT-5.5 | $5 | $0.50 | $30 |
| GPT-5.4 | $2.50 | $0.25 | $15 |
| GPT-5.3 Codex · 5.2 | $1.75 | $0.175 | $14 |
| GPT-5.1 · GPT-5 | $1.25 | $0.125 | $10 |
Checked 2026-10-15 from Anthropic's pricing page. Batch discounts don't apply to Claude Code. Anthropic pricing · OpenAI pricing · Claude plans
FAQ
How do I check my Claude Code usage?
Inside Claude Code, /usage (or /cost on the API) shows the current session and plan limits. To see history by day, project and model, open your ~/.claude/projects folder with this page.
Is my code or chat uploaded?
No. The page reads the files with your browser's File API and only adds up the usage numbers. Your prompts and code never leave the tab.
Why is the API-equivalent cost so high on a Max plan?
Claude Code re-reads your project on every turn, so it moves a lot of cached tokens. On a plan you pay a flat fee; this shows what the same tokens would cost if you paid per token.
Does it work like ccusage or CodeBurn?
It reads the same local files those command-line tools read, but runs in the browser with no install, and adds the plan-vs-API comparison.
Why does it show only the last month?
Claude Code removes logs older than 30 days unless you raise cleanupPeriodDays in settings.json.
Can I check Codex or Cursor usage too?
Yes. For Codex pick ~/.codex/sessions. For Cursor, export the usage CSV from cursor.com/dashboard (Usage → Export CSV) and choose that file. You can load all three at once.
Related: Claude Code monthly cost calculator · Claude Code usage limits explained · Ways to save tokens in Claude Code