Claude Code cost tracker: see what your Pro or Max subscription would cost at API prices, live above your prompt. Spend by model, billing cycle and weekly limit window.
TypeScript
0
10 commits
updated Oct 7, 2026
See what people are saying
README
what-would-it-cost
A Claude Code cost tracker: what your Pro or Max subscription would cost at API prices, live above your prompt.
One line, always on, updated after every model request while Claude works:
| ◉ $4.46 session | this session at API prices. The dot turns orange while Claude is working |
| ▮▮▮▮▮▮ | this session's model mix: orange Opus, purple Fable, blue Sonnet, green Haiku |
| $16.45 today | every Claude Code session on this machine today |
| $315 7d | your current weekly limit window, from Claude Code's own reset time |
| $438 since Sep 28 | your billing cycle so far, from the day your plan renews |
| 2.2x Max 20x | how many times over the cycle has paid for your plan |
| ▂▅█▃▁▆ | the last 14 days, colored from quiet to busy |
| Anthropic is picking up the tab… | a running commentary, with a new line every ten minutes |
It reads your history on install, so the numbers start with your past month instead of $0.
The invoice
Click ≡ invoice on the bar for the full bill. Tabs switch it between this session, your 7-day limit window, your billing cycle and all time, each broken down by model and token type, with a 60-day chart.
Install
In a Claude Code terminal session:
/plugin install what-would-it-cost --marketplace Faouzielbakri/what-would-it-cost
Answer y to add the marketplace, then pick the user scope so it runs in every session. On first run it prices every transcript already on your disk (~/.claude/projects) and tells you the total when it's done.
What it counts
- Every model request, subagents included, priced at Anthropic's list prices per model: input, output, cache reads and cache writes.
- Your history from Claude Code's own transcripts, so nothing is counted twice and nothing needs to be running in the background.
- Prompt caching savings: what the cache reads would have cost as fresh input.
- A cross-check: the session tab shows Claude Code's own
/costtotal next to the mod's. - Every session on this machine for today, 7d and the cycle: open Claude Code in five projects and they all add to the same numbers, updated every 30 seconds.
Privacy and permissions
Everything stays on your machine.
What it sends: nothing. It makes no network requests and has no telemetry. The numbers are kept in Claude Code's own plugin storage as per-day token totals per model.
What it reads:
| What | Why |
|---|---|
Claude Code's transcripts in ~/.claude/projects (or $CLAUDE_CONFIG_DIR/projects) | To price your history on first run. It keeps only each response's model, timestamp and token counts; message text is never stored. |
~/.claude.json, two fields only: oauthAccount.*RateLimitTier and oauthAccount.subscriptionCreatedAt | To know your plan (Pro, Max 5x, Max 20x) and the day it renews. Skipped entirely when you set Your plan and Billing day yourself in /config. |
The HOME and CLAUDE_CONFIG_DIR environment variables | To find the two locations above. |
| Claude Code's own usage figures | For the /cost cross-check and your weekly limit's reset time. |
Programs it runs:
| Command | When and why |
|---|---|
dd if=<transcript> bs=1048576 skip=<n> count=3 | Only for transcripts over 4 MB, which is the most a mod can read in one go. It reads them in 3 MB chunks, read-only. |
claude auth status | Only when the plan can't be read from ~/.claude.json, to tell Pro, Max and API-key accounts apart. |
Your plan
It reads your exact plan (Pro, Max 5x or Max 20x) from the rate-limit tier in Claude Code's own config (~/.claude.json), falling back to claude auth status. To set it yourself, go to /config → what-would-it-cost › Your plan and pick pro, max-5x, max-20x or api.
The multiplier runs over your billing cycle, not the calendar month: if you subscribed on the 15th, the cycle runs from the 15th to the 15th. The renewal day is read from when your subscription began; set Billing day in /config if yours differs (say, after a plan change).
If you're on an API key, the bill is real money. The band shows anyway.
Prices
List prices per million tokens, from Anthropic's pricing page (October 2026):
| Model | Input | Output | Cache read | Cache write (5m / 1h) |
|---|---|---|---|---|
| Fable 5.1 | $10 | $50 | $0.25 | $12.50 / $20 |
| Fable 5 | $10 | $50 | $1.00 | $12.50 / $20 |
| Opus 5.5 | $4 | $20 | $0.20 | $5 / $8 |
| Opus 5 / 4.5–4.8 | $5 | $25 | $0.50 | $6.25 / $10 |
| Sonnet 5 / 5.5 | $2 | $10 | $0.20 | $2.50 / $4 |
| Sonnet 4.x | $3 | $15 | $0.30 | $3.75 / $6 |
| Haiku 4.5 | $1 | $5 | $0.10 | $1.25 / $2 |
Caveats:
- History is priced exactly, cache writes split into 5-minute (1.25x input) and 1-hour (2x) as each transcript records them. Live usage doesn't carry that split, so live writes count as 1-hour, which is what Claude Code uses.
- Claude Code deletes transcripts after 30 days by default (
cleanupPeriodDays), so history only goes back as far as your transcripts do. - Fast mode isn't detected and is priced as standard.
- A model id the table doesn't know is priced as Opus 5.5 and marked with
?. PRs that add new models tohooks/pricing.tsare welcome. - Days are counted in your local time zone.
Develop
claude --plugin-dir ./what-would-it-cost # run it from a checkout; saving a file hot-reloads it
claude plugin validate ./what-would-it-cost
claude plugin test ./what-would-it-cost
License
MIT. Fork it, remix it, ship it. If you build on it, a link back to Faouzi El Bakri is appreciated.
Similar repos
Claude Code cost tracker: see what your Pro or Max subscription would cost at API prices, live above your prompt. Spend by model, billing cycle and weekly limit window.
TypeScript
0
10 commits
updated Oct 7, 2026
See what people are saying
README
what-would-it-cost
A Claude Code cost tracker: what your Pro or Max subscription would cost at API prices, live above your prompt.
One line, always on, updated after every model request while Claude works:
| ◉ $4.46 session | this session at API prices. The dot turns orange while Claude is working |
| ▮▮▮▮▮▮ | this session's model mix: orange Opus, purple Fable, blue Sonnet, green Haiku |
| $16.45 today | every Claude Code session on this machine today |
| $315 7d | your current weekly limit window, from Claude Code's own reset time |
| $438 since Sep 28 | your billing cycle so far, from the day your plan renews |
| 2.2x Max 20x | how many times over the cycle has paid for your plan |
| ▂▅█▃▁▆ | the last 14 days, colored from quiet to busy |
| Anthropic is picking up the tab… | a running commentary, with a new line every ten minutes |
It reads your history on install, so the numbers start with your past month instead of $0.
The invoice
Click ≡ invoice on the bar for the full bill. Tabs switch it between this session, your 7-day limit window, your billing cycle and all time, each broken down by model and token type, with a 60-day chart.
Install
In a Claude Code terminal session:
/plugin install what-would-it-cost --marketplace Faouzielbakri/what-would-it-cost
Answer y to add the marketplace, then pick the user scope so it runs in every session. On first run it prices every transcript already on your disk (~/.claude/projects) and tells you the total when it's done.
What it counts
- Every model request, subagents included, priced at Anthropic's list prices per model: input, output, cache reads and cache writes.
- Your history from Claude Code's own transcripts, so nothing is counted twice and nothing needs to be running in the background.
- Prompt caching savings: what the cache reads would have cost as fresh input.
- A cross-check: the session tab shows Claude Code's own
/costtotal next to the mod's. - Every session on this machine for today, 7d and the cycle: open Claude Code in five projects and they all add to the same numbers, updated every 30 seconds.
Privacy and permissions
Everything stays on your machine.
What it sends: nothing. It makes no network requests and has no telemetry. The numbers are kept in Claude Code's own plugin storage as per-day token totals per model.
What it reads:
| What | Why |
|---|---|
Claude Code's transcripts in ~/.claude/projects (or $CLAUDE_CONFIG_DIR/projects) | To price your history on first run. It keeps only each response's model, timestamp and token counts; message text is never stored. |
~/.claude.json, two fields only: oauthAccount.*RateLimitTier and oauthAccount.subscriptionCreatedAt | To know your plan (Pro, Max 5x, Max 20x) and the day it renews. Skipped entirely when you set Your plan and Billing day yourself in /config. |
The HOME and CLAUDE_CONFIG_DIR environment variables | To find the two locations above. |
| Claude Code's own usage figures | For the /cost cross-check and your weekly limit's reset time. |
Programs it runs:
| Command | When and why |
|---|---|
dd if=<transcript> bs=1048576 skip=<n> count=3 | Only for transcripts over 4 MB, which is the most a mod can read in one go. It reads them in 3 MB chunks, read-only. |
claude auth status | Only when the plan can't be read from ~/.claude.json, to tell Pro, Max and API-key accounts apart. |
Your plan
It reads your exact plan (Pro, Max 5x or Max 20x) from the rate-limit tier in Claude Code's own config (~/.claude.json), falling back to claude auth status. To set it yourself, go to /config → what-would-it-cost › Your plan and pick pro, max-5x, max-20x or api.
The multiplier runs over your billing cycle, not the calendar month: if you subscribed on the 15th, the cycle runs from the 15th to the 15th. The renewal day is read from when your subscription began; set Billing day in /config if yours differs (say, after a plan change).
If you're on an API key, the bill is real money. The band shows anyway.
Prices
List prices per million tokens, from Anthropic's pricing page (October 2026):
| Model | Input | Output | Cache read | Cache write (5m / 1h) |
|---|---|---|---|---|
| Fable 5.1 | $10 | $50 | $0.25 | $12.50 / $20 |
| Fable 5 | $10 | $50 | $1.00 | $12.50 / $20 |
| Opus 5.5 | $4 | $20 | $0.20 | $5 / $8 |
| Opus 5 / 4.5–4.8 | $5 | $25 | $0.50 | $6.25 / $10 |
| Sonnet 5 / 5.5 | $2 | $10 | $0.20 | $2.50 / $4 |
| Sonnet 4.x | $3 | $15 | $0.30 | $3.75 / $6 |
| Haiku 4.5 | $1 | $5 | $0.10 | $1.25 / $2 |
Caveats:
- History is priced exactly, cache writes split into 5-minute (1.25x input) and 1-hour (2x) as each transcript records them. Live usage doesn't carry that split, so live writes count as 1-hour, which is what Claude Code uses.
- Claude Code deletes transcripts after 30 days by default (
cleanupPeriodDays), so history only goes back as far as your transcripts do. - Fast mode isn't detected and is priced as standard.
- A model id the table doesn't know is priced as Opus 5.5 and marked with
?. PRs that add new models tohooks/pricing.tsare welcome. - Days are counted in your local time zone.
Develop
claude --plugin-dir ./what-would-it-cost # run it from a checkout; saving a file hot-reloads it
claude plugin validate ./what-would-it-cost
claude plugin test ./what-would-it-cost
License
MIT. Fork it, remix it, ship it. If you build on it, a link back to Faouzi El Bakri is appreciated.