[BUG]

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

I am reporting a usage-accounting issue on a Claude Max 5x account.

On May 10, 2026, around 17:00 Asia/Tbilisi time, I logged into Claude only to start a new session. I had not used Claude since the previous night. However, the usage page showed the current 5-hour/session window as already fully exhausted.

At the same time, the weekly usage page showed unexplained Sonnet-only usage, even though I intentionally use Opus and do not intentionally use Sonnet.

This appears similar to other Max usage-counter/rate-limit issues reported in this repository, where limits are reached despite little/no corresponding user activity.

Plan

Claude Max 5x

Environment

• OS: Windows
• Shell: PowerShell
• Claude Desktop installed: yes
• Claude Code installed: yes
• Claude PowerPoint/Office add-in installed: yes
• Cowork service present locally: yes, but I have never intentionally used Claude Cowork
• Timezone: Asia/Tbilisi, UTC+4

Actual behavior

On the Claude usage page, I saw:

• Current session: 100% used
• Weekly “All models”: 87% used
• Weekly “Sonnet only”: 39% used
• Claude Design: 0% used
• Daily included routine runs: 0/15

The current session was already exhausted even though I had not used Claude during that session window.

The Sonnet-only usage was also unexpected because I intentionally use Opus, not Sonnet.

Expected behavior

If I have not used Claude during the 5-hour/session window, the current session usage should not be exhausted.

If I use Opus, usage should not appear as unexplained Sonnet-only quota consumption unless there is documented routing or accounting behavior explaining it.

Timeline

• May 9, 2026: I used Claude during the day/night.
• May 10, 2026, around 00:39 Asia/Tbilisi: last local Claude model activity I could find in Claude Code project history.
• May 10, 2026, around 17:00 Asia/Tbilisi: I logged in only to start a new session.
• Immediately around that login, Claude showed my current session as 100% used / limit reached.
• The weekly usage page also showed Sonnet-only usage at 39%.

Local investigation performed

I checked local Claude-related activity from May 9, 2026 00:00 through May 10, 2026 17:00 Asia/Tbilisi.

Findings:

• Claude Code project history showed assistant model activity overwhelmingly under claude-opus-4-7.
• I found no Sonnet-labeled assistant model events in the scanned local Claude Code project history.
• The last local Claude model activity I found before the 17:00 login was around May 10, 2026 00:39 local time.
• I found no local Claude Code project history writes or completion/chat API activity explaining usage between around 00:39 and my login around 17:00.
• Claude Desktop/Cowork logs showed VM/session/OAuth activity, but I found no evidence of a user-started Cowork task or completion activity explaining the 5-hour/session limit exhaustion.
• I have never intentionally used Claude Cowork. Cowork was checked only because support suggested it as a possible source of unexplained usage.
• The Claude PowerPoint/Office add-in is installed, but local checks found no evidence that it made completion requests during the disputed period.
• Browser extension checks did not find a separate Claude/Anthropic browser extension causing usage.
• The /bug command does not work on my side, so I am filing this GitHub issue instead.

Support context

Anthropic support stated that:

• Request-level usage metadata is not available to individual Max subscribers.
• Session-based and weekly usage limits cannot be reset or corrected by support.
• Support cannot escalate internally beyond /bug or /feedback.
• There is no specific documented usage-dispute process beyond /bug.
• A previous support claim about automatic Opus-to-Sonnet switching was later corrected as unsupported by official documentation.

Because /bug does not work for me, there appears to be no working path to get this investigated except filing here.

Why this looks like an accounting/routing issue

There are two related discrepancies:

1. The 5-hour/session limit was exhausted even though I had not used Claude during that session window.
2. The Sonnet-only weekly bucket showed 39% usage even though my local Claude Code history does not show corresponding Sonnet usage and I intentionally use Opus.

Please investigate whether usage is being incorrectly attributed due to one of the following:

• Opus usage being charged to the Sonnet-only bucket
• Claude Desktop session restoration
• Cowork background service activity
• Office/PowerPoint add-in activity
• internal model routing
• stale or duplicated session accounting
• delayed usage reconciliation
• another backend usage-ledger issue

Related public reports

Possibly related issues:

• https://github.com/anthropics/claude-code/issues/39649
• https://github.com/anthropics/claude-code/issues/28832
• https://github.com/anthropics/claude-code/issues/24335
• https://github.com/anthropics/claude-code/issues/41930

Request

Please investigate the server-side usage ledger for this account/session window and determine why:

1. The current 5-hour/session limit was exhausted without corresponding user activity.
2. The Sonnet-only weekly bucket showed 39% usage despite no corresponding local Sonnet usage evidence.

What Should Happen?

I am reporting a usage-accounting issue on a Claude Max 5x account.

On May 10, 2026, around 17:00 Asia/Tbilisi time, I logged into Claude only to start a new session. I had not used Claude since the previous night. However, the usage page showed the current 5-hour/session window as already fully exhausted.

At the same time, the weekly usage page showed unexplained Sonnet-only usage, even though I intentionally use Opus and do not intentionally use Sonnet.

This appears similar to other Max usage-counter/rate-limit issues reported in this repository, where limits are reached despite little/no corresponding user activity.

Plan

Claude Max 5x

Environment

• OS: Windows
• Shell: PowerShell
• Claude Desktop installed: yes
• Claude Code installed: yes
• Claude PowerPoint/Office add-in installed: yes
• Cowork service present locally: yes, but I have never intentionally used Claude Cowork
• Timezone: Asia/Tbilisi, UTC+4

Actual behavior

On the Claude usage page, I saw:

• Current session: 100% used
• Weekly “All models”: 87% used
• Weekly “Sonnet only”: 39% used
• Claude Design: 0% used
• Daily included routine runs: 0/15

The current session was already exhausted even though I had not used Claude during that session window.

The Sonnet-only usage was also unexpected because I intentionally use Opus, not Sonnet.

Expected behavior

If I have not used Claude during the 5-hour/session window, the current session usage should not be exhausted.

If I use Opus, usage should not appear as unexplained Sonnet-only quota consumption unless there is documented routing or accounting behavior explaining it.

Timeline

• May 9, 2026: I used Claude during the day/night.
• May 10, 2026, around 00:39 Asia/Tbilisi: last local Claude model activity I could find in Claude Code project history.
• May 10, 2026, around 17:00 Asia/Tbilisi: I logged in only to start a new session.
• Immediately around that login, Claude showed my current session as 100% used / limit reached.
• The weekly usage page also showed Sonnet-only usage at 39%.

Local investigation performed

I checked local Claude-related activity from May 9, 2026 00:00 through May 10, 2026 17:00 Asia/Tbilisi.

Findings:

• Claude Code project history showed assistant model activity overwhelmingly under claude-opus-4-7.
• I found no Sonnet-labeled assistant model events in the scanned local Claude Code project history.
• The last local Claude model activity I found before the 17:00 login was around May 10, 2026 00:39 local time.
• I found no local Claude Code project history writes or completion/chat API activity explaining usage between around 00:39 and my login around 17:00.
• Claude Desktop/Cowork logs showed VM/session/OAuth activity, but I found no evidence of a user-started Cowork task or completion activity explaining the 5-hour/session limit exhaustion.
• I have never intentionally used Claude Cowork. Cowork was checked only because support suggested it as a possible source of unexplained usage.
• The Claude PowerPoint/Office add-in is installed, but local checks found no evidence that it made completion requests during the disputed period.
• Browser extension checks did not find a separate Claude/Anthropic browser extension causing usage.
• The /bug command does not work on my side, so I am filing this GitHub issue instead.

Support context

Anthropic support stated that:

• Request-level usage metadata is not available to individual Max subscribers.
• Session-based and weekly usage limits cannot be reset or corrected by support.
• Support cannot escalate internally beyond /bug or /feedback.
• There is no specific documented usage-dispute process beyond /bug.
• A previous support claim about automatic Opus-to-Sonnet switching was later corrected as unsupported by official documentation.

Because /bug does not work for me, there appears to be no working path to get this investigated except filing here.

Why this looks like an accounting/routing issue

There are two related discrepancies:

1. The 5-hour/session limit was exhausted even though I had not used Claude during that session window.
2. The Sonnet-only weekly bucket showed 39% usage even though my local Claude Code history does not show corresponding Sonnet usage and I intentionally use Opus.

Please investigate whether usage is being incorrectly attributed due to one of the following:

• Opus usage being charged to the Sonnet-only bucket
• Claude Desktop session restoration
• Cowork background service activity
• Office/PowerPoint add-in activity
• internal model routing
• stale or duplicated session accounting
• delayed usage reconciliation
• another backend usage-ledger issue

Related public reports

Possibly related issues:

• https://github.com/anthropics/claude-code/issues/39649
• https://github.com/anthropics/claude-code/issues/28832
• https://github.com/anthropics/claude-code/issues/24335
• https://github.com/anthropics/claude-code/issues/41930

Request

Please investigate the server-side usage ledger for this account/session window and determine why:

1. The current 5-hour/session limit was exhausted without corresponding user activity.
2. The Sonnet-only weekly bucket showed 39% usage despite no corresponding local Sonnet usage evidence.

Error Messages/Logs

Steps to Reproduce

I am reporting a usage-accounting issue on a Claude Max 5x account.

On May 10, 2026, around 17:00 Asia/Tbilisi time, I logged into Claude only to start a new session. I had not used Claude since the previous night. However, the usage page showed the current 5-hour/session window as already fully exhausted.

At the same time, the weekly usage page showed unexplained Sonnet-only usage, even though I intentionally use Opus and do not intentionally use Sonnet.

This appears similar to other Max usage-counter/rate-limit issues reported in this repository, where limits are reached despite little/no corresponding user activity.

Plan

Claude Max 5x

Environment

• OS: Windows
• Shell: PowerShell
• Claude Desktop installed: yes
• Claude Code installed: yes
• Claude PowerPoint/Office add-in installed: yes
• Cowork service present locally: yes, but I have never intentionally used Claude Cowork
• Timezone: Asia/Tbilisi, UTC+4

Actual behavior

On the Claude usage page, I saw:

• Current session: 100% used
• Weekly “All models”: 87% used
• Weekly “Sonnet only”: 39% used
• Claude Design: 0% used
• Daily included routine runs: 0/15

The current session was already exhausted even though I had not used Claude during that session window.

The Sonnet-only usage was also unexpected because I intentionally use Opus, not Sonnet.

Expected behavior

If I have not used Claude during the 5-hour/session window, the current session usage should not be exhausted.

If I use Opus, usage should not appear as unexplained Sonnet-only quota consumption unless there is documented routing or accounting behavior explaining it.

Timeline

• May 9, 2026: I used Claude during the day/night.
• May 10, 2026, around 00:39 Asia/Tbilisi: last local Claude model activity I could find in Claude Code project history.
• May 10, 2026, around 17:00 Asia/Tbilisi: I logged in only to start a new session.
• Immediately around that login, Claude showed my current session as 100% used / limit reached.
• The weekly usage page also showed Sonnet-only usage at 39%.

Local investigation performed

I checked local Claude-related activity from May 9, 2026 00:00 through May 10, 2026 17:00 Asia/Tbilisi.

Findings:

• Claude Code project history showed assistant model activity overwhelmingly under claude-opus-4-7.
• I found no Sonnet-labeled assistant model events in the scanned local Claude Code project history.
• The last local Claude model activity I found before the 17:00 login was around May 10, 2026 00:39 local time.
• I found no local Claude Code project history writes or completion/chat API activity explaining usage between around 00:39 and my login around 17:00.
• Claude Desktop/Cowork logs showed VM/session/OAuth activity, but I found no evidence of a user-started Cowork task or completion activity explaining the 5-hour/session limit exhaustion.
• I have never intentionally used Claude Cowork. Cowork was checked only because support suggested it as a possible source of unexplained usage.
• The Claude PowerPoint/Office add-in is installed, but local checks found no evidence that it made completion requests during the disputed period.
• Browser extension checks did not find a separate Claude/Anthropic browser extension causing usage.
• The /bug command does not work on my side, so I am filing this GitHub issue instead.

Support context

Anthropic support stated that:

• Request-level usage metadata is not available to individual Max subscribers.
• Session-based and weekly usage limits cannot be reset or corrected by support.
• Support cannot escalate internally beyond /bug or /feedback.
• There is no specific documented usage-dispute process beyond /bug.
• A previous support claim about automatic Opus-to-Sonnet switching was later corrected as unsupported by official documentation.

Because /bug does not work for me, there appears to be no working path to get this investigated except filing here.

Why this looks like an accounting/routing issue

There are two related discrepancies:

1. The 5-hour/session limit was exhausted even though I had not used Claude during that session window.
2. The Sonnet-only weekly bucket showed 39% usage even though my local Claude Code history does not show corresponding Sonnet usage and I intentionally use Opus.

Please investigate whether usage is being incorrectly attributed due to one of the following:

• Opus usage being charged to the Sonnet-only bucket
• Claude Desktop session restoration
• Cowork background service activity
• Office/PowerPoint add-in activity
• internal model routing
• stale or duplicated session accounting
• delayed usage reconciliation
• another backend usage-ledger issue

Related public reports

Possibly related issues:

• https://github.com/anthropics/claude-code/issues/39649
• https://github.com/anthropics/claude-code/issues/28832
• https://github.com/anthropics/claude-code/issues/24335
• https://github.com/anthropics/claude-code/issues/41930

Request

Please investigate the server-side usage ledger for this account/session window and determine why:

1. The current 5-hour/session limit was exhausted without corresponding user activity.
2. The Sonnet-only weekly bucket showed 39% usage despite no corresponding local Sonnet usage evidence.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

I am reporting a usage-accounting issue on a Claude Max 5x account. On May 10, 2026, around 17:00 Asia/Tbilisi time, I logged into Claude only to start a new session. I had not used Claude since the previous night. However, the usage page showed the current 5-hour/session window as already fully exhausted. At the same time, the weekly usage page showed unexplained Sonnet-only usage, even though I intentionally use Opus and do not intentionally use Sonnet. This appears similar to other Max usage-counter/rate-limit issues reported in this repository, where limits are reached despite little/no corresponding user activity. ## Plan Claude Max 5x ## Environment - OS: Windows - Shell: PowerShell - Claude Desktop installed: yes - Claude Code installed: yes - Claude PowerPoint/Office add-in installed: yes - Cowork service present locally: yes, but I have never intentionally used Claude Cowork - Timezone: Asia/Tbilisi, UTC+4 ## Actual behavior On the Claude usage page, I saw: - Current session: 100% used - Weekly “All models”: 87% used - Weekly “Sonnet only”: 39% used - Claude Design: 0% used - Daily included routine runs: 0/15 The current session was already exhausted even though I had not used Claude during that session window. The Sonnet-only usage was also unexpected because I intentionally use Opus, not Sonnet. ## Expected behavior If I have not used Claude during the 5-hour/session window, the current session usage should not be exhausted. If I use Opus, usage should not appear as unexplained Sonnet-only quota consumption unless there is documented routing or accounting behavior explaining it. ## Timeline - May 9, 2026: I used Claude during the day/night. - May 10, 2026, around 00:39 Asia/Tbilisi: last local Claude model activity I could find in Claude Code project history. - May 10, 2026, around 17:00 Asia/Tbilisi: I logged in only to start a new session. - Immediately around that login, Claude showed my current session as 100% used / limit reached. - The weekly usage page also showed Sonnet-only usage at 39%. ## Local investigation performed I checked local Claude-related activity from May 9, 2026 00:00 through May 10, 2026 17:00 Asia/Tbilisi. Findings: - Claude Code project history showed assistant model activity overwhelmingly under claude-opus-4-7. - I found no Sonnet-labeled assistant model events in the scanned local Claude Code project history. - The last local Claude model activity I found before the 17:00 login was around May 10, 2026 00:39 local time. - I found no local Claude Code project history writes or completion/chat API activity explaining usage between around 00:39 and my login around 17:00. - Claude Desktop/Cowork logs showed VM/session/OAuth activity, but I found no evidence of a user-started Cowork task or completion activity explaining the 5-hour/session limit exhaustion. - I have never intentionally used Claude Cowork. Cowork was checked only because support suggested it as a possible source of unexplained usage. - The Claude PowerPoint/Office add-in is installed, but local checks found no evidence that it made completion requests during the disputed period. - Browser extension checks did not find a separate Claude/Anthropic browser extension causing usage. - The /bug command does not work on my side, so I am filing this GitHub issue instead. ## Support context Anthropic support stated that: - Request-level usage metadata is not available to individual Max subscribers. - Session-based and weekly usage limits cannot be reset or corrected by support. - Support cannot escalate internally beyond /bug or /feedback. - There is no specific documented usage-dispute process beyond /bug. - A previous support claim about automatic Opus-to-Sonnet switching was later corrected as unsupported by official documentation. Because /bug does not work for me, there appears to be no working path to get this investigated except filing here. ## Why this looks like an accounting/routing issue There are two related discrepancies: 1. The 5-hour/session limit was exhausted even though I had not used Claude during that session window. 2. The Sonnet-only weekly bucket showed 39% usage even though my local Claude Code history does not show corresponding Sonnet usage and I intentionally use Opus. Please investigate whether usage is being incorrectly attributed due to one of the following: - Opus usage being charged to the Sonnet-only bucket - Claude Desktop session restoration - Cowork background service activity - Office/PowerPoint add-in activity - internal model routing - stale or duplicated session accounting - delayed usage reconciliation - another backend usage-ledger issue ## Related public reports Possibly related issues: - https://github.com/anthropics/claude-code/issues/39649 - https://github.com/anthropics/claude-code/issues/28832 - https://github.com/anthropics/claude-code/issues/24335 - https://github.com/anthropics/claude-code/issues/41930 ## Request Please investigate the server-side usage ledger for this account/session window and determine why: 1. The current 5-hour/session limit was exhausted without corresponding user activity. 2. The Sonnet-only weekly bucket showed 39% usage despite no corresponding local Sonnet usage evidence.

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

I am reporting a usage-accounting issue on a Claude Max 5x account.

On May 10, 2026, around 17:00 Asia/Tbilisi time, I logged into Claude only to start a new session. I had not used Claude since the previous night. However, the usage page showed the current 5-hour/session window as already fully exhausted.

At the same time, the weekly usage page showed unexplained Sonnet-only usage, even though I intentionally use Opus and do not intentionally use Sonnet.

This appears similar to other Max usage-counter/rate-limit issues reported in this repository, where limits are reached despite little/no corresponding user activity.

Plan

Claude Max 5x

Environment

• OS: Windows
• Shell: PowerShell
• Claude Desktop installed: yes
• Claude Code installed: yes
• Claude PowerPoint/Office add-in installed: yes
• Cowork service present locally: yes, but I have never intentionally used Claude Cowork
• Timezone: Asia/Tbilisi, UTC+4

Actual behavior

On the Claude usage page, I saw:

• Current session: 100% used
• Weekly “All models”: 87% used
• Weekly “Sonnet only”: 39% used
• Claude Design: 0% used
• Daily included routine runs: 0/15

The current session was already exhausted even though I had not used Claude during that session window.

The Sonnet-only usage was also unexpected because I intentionally use Opus, not Sonnet.

Expected behavior

If I have not used Claude during the 5-hour/session window, the current session usage should not be exhausted.

If I use Opus, usage should not appear as unexplained Sonnet-only quota consumption unless there is documented routing or accounting behavior explaining it.

Timeline

• May 9, 2026: I used Claude during the day/night.
• May 10, 2026, around 00:39 Asia/Tbilisi: last local Claude model activity I could find in Claude Code project history.
• May 10, 2026, around 17:00 Asia/Tbilisi: I logged in only to start a new session.
• Immediately around that login, Claude showed my current session as 100% used / limit reached.
• The weekly usage page also showed Sonnet-only usage at 39%.

Local investigation performed

I checked local Claude-related activity from May 9, 2026 00:00 through May 10, 2026 17:00 Asia/Tbilisi.

Findings:

• Claude Code project history showed assistant model activity overwhelmingly under claude-opus-4-7.
• I found no Sonnet-labeled assistant model events in the scanned local Claude Code project history.
• The last local Claude model activity I found before the 17:00 login was around May 10, 2026 00:39 local time.
• I found no local Claude Code project history writes or completion/chat API activity explaining usage between around 00:39 and my login around 17:00.
• Claude Desktop/Cowork logs showed VM/session/OAuth activity, but I found no evidence of a user-started Cowork task or completion activity explaining the 5-hour/session limit exhaustion.
• I have never intentionally used Claude Cowork. Cowork was checked only because support suggested it as a possible source of unexplained usage.
• The Claude PowerPoint/Office add-in is installed, but local checks found no evidence that it made completion requests during the disputed period.
• Browser extension checks did not find a separate Claude/Anthropic browser extension causing usage.
• The /bug command does not work on my side, so I am filing this GitHub issue instead.

Support context

Anthropic support stated that:

• Request-level usage metadata is not available to individual Max subscribers.
• Session-based and weekly usage limits cannot be reset or corrected by support.
• Support cannot escalate internally beyond /bug or /feedback.
• There is no specific documented usage-dispute process beyond /bug.
• A previous support claim about automatic Opus-to-Sonnet switching was later corrected as unsupported by official documentation.

Because /bug does not work for me, there appears to be no working path to get this investigated except filing here.

Why this looks like an accounting/routing issue

There are two related discrepancies:

1. The 5-hour/session limit was exhausted even though I had not used Claude during that session window.
2. The Sonnet-only weekly bucket showed 39% usage even though my local Claude Code history does not show corresponding Sonnet usage and I intentionally use Opus.

Please investigate whether usage is being incorrectly attributed due to one of the following:

• Opus usage being charged to the Sonnet-only bucket
• Claude Desktop session restoration
• Cowork background service activity
• Office/PowerPoint add-in activity
• internal model routing
• stale or duplicated session accounting
• delayed usage reconciliation
• another backend usage-ledger issue

Related public reports

Possibly related issues:

• https://github.com/anthropics/claude-code/issues/39649
• https://github.com/anthropics/claude-code/issues/28832
• https://github.com/anthropics/claude-code/issues/24335
• https://github.com/anthropics/claude-code/issues/41930

Request

Please investigate the server-side usage ledger for this account/session window and determine why:

1. The current 5-hour/session limit was exhausted without corresponding user activity.
2. The Sonnet-only weekly bucket showed 39% usage despite no corresponding local Sonnet usage evidence.