The water estimate

This is the playful part of cca, and the part where being honest matters most.

There is no reliable figure

No trustworthy public per-token water figure exists for any model. Published estimates vary by more than an order of magnitude depending on:

• where the datacenter is, and what its climate demands of cooling;
• the cooling design — evaporative towers consume far more water than closed-loop or air-cooled systems;
• the power mix, since thermoelectric generation consumes water too;
• and crucially, whether the figure counts only on-site evaporation or also the water consumed generating the electricity.

Anyone quoting a precise number is extrapolating from one datacenter, one cooling design, and one set of accounting boundaries.

So cca does the least-bad thing

One constant — 0.30 mL per 1,000 tokens — applied flat across every model and every token class, with the assumption printed beside the result every single time:

Water   ≈ 18.0 L   about 2 minutes of shower

─ Water assumes 0.30 mL / 1k tokens, a rough estimate. See README.

The figure is never shown without the caveat that produced it.

It is a placeholder, not a measurement

Treat the water figure as an order-of-magnitude illustration. It is there to make an abstract quantity feel concrete, not to support a claim.

Why it is flat

Scaling per model or per token class would be false precision dressed up as rigour. The underlying figure is uncertain by more than 10x; differentiating a cache read from an output token to three significant figures would imply a confidence that does not exist.

Substitute your own

If you have a figure you trust more — from your provider, or from research you find credible — use it:

cca --water-ml-per-1k 0.12

Or permanently, in ~/.config/cca/config.json:

{ "water_ml_per_1k_tokens": 0.12 }

The printed assumption updates to match, so the output never claims a rate it did not use.

Per-session water

cca sessions shows each session’s share:

  Claude Code usage · by session · all time

  Title                                Session    water  tokens   cost
  refactor: split the billing service  7c1a9f20   2.1 L    7.2M  $9.30
  feat: dashboard charts and filters   9c1a77b0   1.8 L    6.0M  $8.20
  Invoice rounding bug investigation   3ef0d219   1.6 L    5.2M  $7.01
  —                                    51f4fc3e  841 mL    2.8M  $3.42

  ─ Cost is what this usage would cost at API rates, not what you were billed;
    Claude Code on a subscription draws from your plan allowance instead.
  ─ Water assumes 0.30 mL / 1k tokens, a rough estimate. See README.

The column is the token column rescaled

Water is a flat rate per token, so a session’s water figure is exactly its token total in different units. Ranking sessions by water gives the same order as ranking by tokens, and the ratio between any two rows is identical in both columns.

That is the point rather than a flaw — “2.1 L” means something to a person in a way “7.2M tokens” does not. But the column reveals nothing the token column did not, and nothing in cca sorts, filters, or aggregates by it.

Rows carry the volume alone. An equivalence is what makes a single figure concrete; twenty of them down a column would be noise, so those stay on the summary.

--no-water hides the column:

  Claude Code usage · by session · all time

  Title                                Session      started  tokens   cost
  refactor: split the billing service  7c1a9f20  2026-09-14    7.2M  $9.30
  feat: dashboard charts and filters   9c1a77b0  2026-09-14    6.0M  $8.20
  Invoice rounding bug investigation   3ef0d219  2026-09-14    5.2M  $7.01
  —                                    51f4fc3e  2026-09-15    2.8M  $3.42

  ─ Cost is what this usage would cost at API rates, not what you were billed;
    Claude Code on a subscription draws from your plan allowance instead.

In --json and --csv the figure is always present, on every grouping, since those are not displays.

The equivalences

The litre figure is converted to whatever familiar unit it covers at least once, which keeps the count small and readable:

Unit	Volume
Cup of coffee	0.25 L
Bottle of water	0.5 L
Minute of shower	9 L
Load of laundry	50 L
Bathtub	150 L
Swimming pool	50,000 L

Counts are rounded to whole numbers, because a decimal would imply precision the estimate does not have.

The equivalences are deliberately lighthearted and carry no judgement. The tool reports; it does not lecture. There is a test that guards the wording against drifting into it.