fuzzyworld-tools / usage

Claude Code usage

Read from the FW-Tokens-* trailers in commit messages (or older Usage: lines).

Usage across all repositories

Commits with usage data

45 of 69

Sessions

17

Output tokens

~3.5M

Cached reads

~885.4M

Estimated cost

est. $21.09 over the 7 commits with a known costestimated API list-price equivalent at prices of 2026-10-22, not what was paid

Output tokens per week

Output tokens of commits with usage trailers; the main measure of work done.

Estimated cost per week

Sum of known per-commit costs: estimated API list-price equivalent, not what was paid. Commits without a known cost are left out, not counted as zero.

Cached reads per week

Tokens read from the prompt cache. They grow with context length and turns, so they track session length more than work done.

Cost per commit, average per week

Known cost divided by the commits with a known cost that week (n is in the tooltip): estimated API list-price equivalent, not what was paid. A commit's cost includes dead ends since the previous trailer.

AI-marked commits per week

Grey line: all commits. Blue line: commits with an AI marker (an AI Co-Authored-By trailer, or a usage trailer). The rest have no AI marker, which does not mean a human wrote them: a marker is only there when the tool added it.

Commits by output tokens

Sessions

How this is measured

• A trailer is the running token tally of one Claude Code session. A commit's cost is the step from the previous commit's tally. A new session starts, and its first commit gets the whole tally, at the first trailer, when the counting method changes, when the output or cached total falls, when any model's counter falls, when a model that had tokens disappears, or when per-model lines appear after commits that had none.
• The bars and rankings in the commit table use output tokens. Cached reads grow with how long the context is and how many turns ran, so they track session length more than the work; input tokens are negligible. These are tokens; dollar figures are estimated API list-price equivalents, not what was paid.
• Costs need the per-model trailers (FW-Model-Tokens) and are estimates at list prices as of 2026-10-22. Commits without them show cost unknown, and unknown is not zero: it is left out of the cost total and drawn hatched in the chart.
• Trailers without FW-Tokens-Method: 2 counted duplicate transcript lines (roughly 1.5-2× too high) and missed subagents; they are shown as written.
• Tallies are rounded to a tenth of K, M or G, so a step is only known to within about two such tenths. Steps at or under that show as ≈0.
• A commit with no trailer shows no bar, and the work before it is counted in the next commit that has one. The cost is of everything done since the previous trailer, including dead ends, not only the code in the commit.
• Highlighted bars in the commit table: at least 2× the median commit.