Claude Code detail breakdown

[ChrisGVE]

Is it my configuration that is wrong or we had a regression (likely due to Claude changes):

Here is what I see:

And here is my provider configuration:

Not so long ago there was a breakdown of the session, the weekly quota, Sonnet. But now I only see a 100% left session which is not correct.

Can you tell me if my configuration is wrong, and what to change, or if there was a regression that is being worked on.

Thanks!

[clawsweeper]

Codex review: keeping this open for maintainer follow-up; there is still a little grit to resolve. Reviewed June 10, 2026, 4:49 AM ET / 08:49 UTC.

Summary
Keep open. The report is current, the screenshots show Claude resolving to the CLI source and rendering only Session: 100% left, and current source has no evidence of a merged fix after the report; the missing raw Claude usage/diagnostic output is the main blocker to choosing the exact parser or fallback repair.

Reproducibility: no. not as a high-confidence end-to-end reproduction. The screenshot plus source inspection reproduce the one-row rendering path when Claude supplies only a primary window, but the raw Claude 2.1.170 /usage output is still needed to confirm the parser failure mode.

Ways to help us reproduce this

• Add expected vs actual behavior.

Next step
Maintainer or reporter input is needed first because the repair depends on the redacted Claude 2.1.170 /usage or codexbar diagnose --provider claude --format json --pretty output, and live provider probes are intentionally out of scope for this review.

Review details

Best possible solution:

Keep this issue open, collect a redacted Claude diagnostic or raw /usage capture for Claude 2.1.170, then land a focused parser/fallback fix that restores weekly and Sonnet lanes without breaking the existing enterprise/web null-window behavior.

Do we have a high-confidence way to reproduce the issue?

No, not as a high-confidence end-to-end reproduction. The screenshot plus source inspection reproduce the one-row rendering path when Claude supplies only a primary window, but the raw Claude 2.1.170 /usage output is still needed to confirm the parser failure mode.

Is this the best way to solve the issue?

Unclear. The likely fix is a narrow Claude CLI parser or source fallback adjustment, but it should be driven by redacted real output and must preserve the existing enterprise/web null-session behavior covered by tests.

AGENTS.md: found and applied where relevant.

Remaining risk / open question:

• The screenshots show the app symptom but not the raw Claude /usage text, so it is unclear whether Claude CLI stopped emitting weekly/Sonnet quota data or CodexBar is failing to parse a renamed/moved panel.
• Changing the null-session web fallback without preserving the enterprise-account tests could regress a prior intentional fix for enterprise/credit-based Claude accounts.

Codex review notes: model gpt-5.5, reasoning high; reviewed against e97bfb0db9eb.

Label changes

Label changes:

• add P2: The report describes a normal-priority provider usage display regression with limited blast radius to Claude Code quota details.
• add impact:other: The issue affects maintainer-visible quota display correctness for Claude usage, which is outside the more specific owned impact categories.
• add issue-rating: 🐚 platinum hermit: Current issue advisory state selects this label.
• add clawsweeper:needs-live-repro: Current issue advisory state selects this label.
• add clawsweeper:no-new-fix-pr: Current issue advisory state selects this label.
• add clawsweeper:needs-maintainer-review: Current issue advisory state selects this label.

Label justifications:

• P2: The report describes a normal-priority provider usage display regression with limited blast radius to Claude Code quota details.
• impact:other: The issue affects maintainer-visible quota display correctness for Claude usage, which is outside the more specific owned impact categories.

Evidence reviewed

Acceptance criteria:

• swift test --filter ClaudeUsageTests
• swift test --filter StatusProbeTests
• swift test --filter ClaudeSourcePlannerTests
• make check

What I checked:

• Issue evidence: The live issue body and attachments show the Claude menu with only a Session bar at 100% left; the settings screenshot shows Usage source set to Auto with resolved source claude, Keychain prompt policy Never prompt, cached Safari Claude cookies, and Claude CLI 2.1.170.
• Current main timing: The checked-out main commit is e97bfb0 from 2026-06-10T07:50:38Z, which predates the issue creation at 2026-06-10T08:41:35Z; no later local main change can be the fix in this checkout. (e97bfb0db9eb)
• Claude source selection: The Claude provider app auto plan orders OAuth, CLI, then Web, and the CLI strategy reports source label claude, matching the reporter's resolved source label in the settings screenshot. (Sources/CodexBarCore/Providers/Claude/ClaudeProviderDescriptor.swift:54, e97bfb0db9eb)
• Claude CLI parser boundary: The CLI parser only populates weekly and Sonnet windows when it can find Current week/Sonnet labels or corresponding ordered percentages, so a changed Claude /usage panel can naturally produce a snapshot with only the primary session window. (Sources/CodexBarCore/Providers/Claude/ClaudeStatusProbe.swift:185, e97bfb0db9eb)
• Menu rendering boundary: The menu and menu card render weekly and Sonnet rows only when snapshot.secondary and snapshot.tertiary exist, explaining why a Claude snapshot missing those windows becomes a one-row Session display. (Sources/CodexBar/MenuDescriptor.swift:197, e97bfb0db9eb)
• Relevant prior fixes: Prior Claude work restored Web-mode Sonnet parsing and handled Claude CLI 2.1 subscription-only output, but those merged fixes are older than this report and do not provide proof that Claude CLI 2.1.170's current output shape is handled.

Likely related people:

• steipete: Recent commits and blame show repeated work on Claude usage fetching, source planning, web usage handling, and menu presentation around the affected code paths. (role: recent area contributor; confidence: high; commits: 4da1e510df8e, c6ad9901a1d4, aa8977ef994d; files: Sources/CodexBarCore/Providers/Claude/ClaudeUsageFetcher.swift, Sources/CodexBarCore/Providers/Claude/ClaudeSourcePlanner.swift, Sources/CodexBarCore/Providers/Claude/ClaudeWeb/ClaudeWebAPIFetcher.swift)
• Yuxin-Qiao: Authored the recent Claude CLI subscription-output handling for Claude CLI 2.1, adjacent to this report's Claude 2.1.170 CLI-source symptoms. (role: recent CLI parser contributor; confidence: medium; commits: 126dc9238715, 11f920652804; files: Sources/CodexBarCore/Providers/Claude/ClaudeStatusProbe.swift, Sources/CodexBarCore/Providers/Claude/ClaudeUsageFetcher.swift)
• clintandrewhall: Authored the enterprise web-account fallback that intentionally maps null five_hour data to 0% used, an adjacent behavior to misleading 100% left displays if the web path is involved. (role: web parser behavior contributor; confidence: medium; commits: 34d9b187a508; files: Sources/CodexBarCore/Providers/Claude/ClaudeWeb/ClaudeWebAPIFetcher.swift, Tests/CodexBarTests/ClaudeWebEnterpriseUsageTests.swift)
• AISupplyGuy: Authored the merged Claude extra usage work that restored Web-mode Sonnet parsing and added related menu metric coverage, directly related to the requested Sonnet breakdown behavior. (role: Claude usage bars feature contributor; confidence: medium; commits: 06e732e29fa1; files: Sources/CodexBarCore/Providers/Claude/ClaudeWeb/ClaudeWebAPIFetcher.swift, Sources/CodexBar/MenuCardView.swift, Tests/CodexBarTests/ClaudeUsageTests.swift)
• ratulsarna: GitHub commit history shows prior source-planner consolidation and compact-label/reset parsing work in the same Claude provider area. (role: Claude source/planner and parser contributor; confidence: medium; commits: 468950d5b8f8, 02bd8e8d0f76; files: Sources/CodexBarCore/Providers/Claude/ClaudeSourcePlanner.swift, Sources/CodexBarCore/Providers/Claude/ClaudeStatusProbe.swift)

How this review workflow works

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.