can you change the status line so I can see the `Usage credits` of the month?

Claude Code·Opus 5.[1m]·nodo·2w ago·29min·1 Checkpoint·24 file changes·+948/-40·17.1K tokens

can you change the status line so I can see the Usage credits of the month?

2w ago·9m·28 calls

Done — ~/.claude/statusline-command.sh now shows month-to-date usage credits:

With a spend cap configured it renders credits $1,034.10/$2,000.00 (52%). Your account has monthly_limit: null (unlimited), so only the spent amount shows. Old script backed up at statusline-command.sh.bak; settings.json already pointed here, so no config change was needed.

Where the number comes from: the statusline JSON Claude Code feeds to the script only carries rate_limits.five_hour / seven_day — no credits field. So the script reads Claude Code's own cached /api/oauth/usage payload from ~/.claude.json (.cachedUsageUtilization.utilization.spend, falling back to .extra_usage for the older schema). No network call, no token handling, ~15 ms per render.

The one caveat: that cache is refreshed only when Claude Code fetches usage (opening /usage, /status, /usage-credits), not on a timer — so it can go stale. Rather than hide that, the script appends the age once it's over 30 minutes old (credits $1,034.10 (2h ago)). Run /usage to force a refresh. Restart or /statusline reload to see it in this session.

fe0c3b5Separate an empty source from an unmatched scope, and let a verified-empty source converge Replicate failed every run whose planning produced no desired refs, with one message covering two unrelated conditions: the source has no refs, and the source has refs that the requested scope excluded. A caller could not tell them apart, and the first is not always a failure — a mirror of a repository that has never been pushed to is trivially up to date, yet it read as an error forever. SyncPolicy.AllowEmptySource (off by default) opts into the distinction. With it set, Replicate reports ErrNoRefsSelected when the source does advertise refs, ErrSourceEmptyUnverified when it advertised none but never confirmed it is empty, ErrSourceEmptyTargetPopulated when it is confirmed empty while the target still holds refs, and a zero-plan success carrying ExecutionSummary.SourceEmpty when source and target are both empty and therefore already agree. Emptiness is established from what the source asserts, never inferred from a response that merely carried no refs. git-sync now requests protocol v2's ls-refs=unborn where the server advertises it, so a repository with no commits answers with an explicit "unborn HEAD" line; only that assertion, under an all-refs scope, qualifies. The distinction is the point: a blank body behind a valid header, a server-side ref-listing or hide-pattern regression, or a narrowed ref-prefix all produce the same silence as an empty repository, and a caller acting on silence would act on every affected repository at once. The unborn line's symref-target is deliberately not surfaced as SourceHEAD, which consumers read as a branch that exists. The divergent case refuses rather than converging. Converging means deleting every ref on the target, and the states that produce that signature — a source restored from backup, a wiped data plane, an out-of-band emptying — are the ones where the target may hold the only surviving copy. The opt-in gate is checked first, so "off" is structurally identical to the behavior that predates this and not merely identical in the cases someone thought to test: a caller that has not opted in cannot receive a sentinel it has never heard of. The sentinels' messages deliberately avoid the historical "no source refs matched" text, so a caller that substring-matches that phrase cannot read one as the other and the order the checks run in is not load-bearing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01714HJZAqpgwuwp6fcMWEhG Entire-Checkpoint: 01M0JBEEG33N57NPRZ1MAT4DY6+468/-18
d71f306Separate an empty source from an unmatched scope, and let a verified-empty source converge Replicate failed every run whose planning produced no desired refs, with one message covering two unrelated conditions: the source has no refs, and the source has refs that the requested scope excluded. A caller could not tell them apart, and the first is not always a failure — a mirror of a repository that has never been pushed to is trivially up to date, yet it read as an error forever. SyncPolicy.AllowEmptySource (off by default) opts into the distinction. With it set, Replicate reports ErrNoRefsSelected when the source does advertise refs, ErrSourceEmptyUnverified when it advertised none but never confirmed it is empty, ErrSourceEmptyTargetPopulated when it is confirmed empty while the target still holds refs, and a zero-plan success carrying ExecutionSummary.SourceEmpty when source and target are both empty and therefore already agree. Emptiness is established from what the source asserts, never inferred from a response that merely carried no refs. git-sync now requests protocol v2's ls-refs=unborn where the server advertises it, so a repository with no commits answers with an explicit "unborn HEAD" line; only that assertion, under an all-refs scope, qualifies. The distinction is the point: a blank body behind a valid header, a server-side ref-listing or hide-pattern regression, or a narrowed ref-prefix all produce the same silence as an empty repository, and a caller acting on silence would act on every affected repository at once. The unborn line's symref-target is deliberately not surfaced as SourceHEAD, which consumers read as a branch that exists. The divergent case refuses rather than converging. Converging means deleting every ref on the target, and the states that produce that signature — a source restored from backup, a wiped data plane, an out-of-band emptying — are the ones where the target may hold the only surviving copy. The opt-in gate is checked first, so "off" is structurally identical to the behavior that predates this and not merely identical in the cases someone thought to test: a caller that has not opted in cannot receive a sentinel it has never heard of. The sentinels' messages deliberately avoid the historical "no source refs matched" text, so a caller that substring-matches that phrase cannot read one as the other and the order the checks run in is not load-bearing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01714HJZAqpgwuwp6fcMWEhG Entire-Checkpoint: 01M0JBEEG33N57NPRZ1MAT4DY6+480/-22