Billing reconciliation matrix

Map the symptoms to the Anthropic systems that need owner review.

A privacy-sanitized matrix for reporters, support reviewers, and a human Anthropic Billing Platform / payment-operations owner. It separates the observed customer failure from related public patterns so the case is not dismissed as a generic card decline.

Updated: 2026-07-27 15:47 JST
Status: unresolved based on current evidence
Scope: Claude usage-credit billing and entitlement reconciliation

Core Contradiction

Manual prepaid Claude usage-credit purchases fail or lose the advertised discounted checkout state, while automatic usage-credit top-ups from the same saved billing setup continue charging successfully. A support-route readback at 2026-07-27 03:16 JST found one official support thread closed after generic card-decline bot guidance and a later thread where Fin acknowledged a billing-system technical contradiction, offered human connection, then created a new pending support conversation, with no confirmed human owner or restored manual purchase path. That contradiction needs cross-system reconciliation, not another generic update-payment-method answer.

Latest Public Route Refresh

Between 2026-07-26 22:37 and 2026-07-27 16:31 JST, existing Anthropic-owned GitHub comments were refreshed or extended in place with the 2026-07-26 22:26 JST support posture, later verified independent guide corroboration, including a newly verified Claude Code pricing-guide caveat, a fresh paid-account usage-meter/support-routing report, a session-limit / zero-token usage-meter mismatch report, a silent cross-account billing/auth route, a central billing-pattern anchor tying the guide corroboration to the #81116/#72552 route evidence, additional public Reddit payment-initiation / credit-ledger, invoice-lock, support-loop, gift-entitlement, and hidden billing-route reports, plus public upgrade-payment, paid-invoice/past-due, failed-payment-notification, refund-credit-state, large-billing-error, refund-window, known-billing-system-error/manual-intervention, mid-cycle-downgrade/double-charge, unauthorized-plan-mutation, gift-cancellation/redemption-failure, unauthorized-gift-charge/support-loss, support-loop reports, a new public X reply on a high-visibility official Claude launch thread, a directly relevant Facebook Claude Community billing/support-thread comment, newly added public subscription/extra-usage billing-mode reports, fresh Anthropic-owned cost-control comments covering visible-ledger/overage enforcement mismatch, wrong monthly-spend remediation, paid-work loss after spend-limit interruption, orphaned background-agent work after ambiguous limiter failures, recursive agent cost-burn guardrails, a fresh paid-tier allocation / weekly-limit reconciliation route, phantom-usage / idle quota-drain / usage-leak cost-control routes, a billing-source / scheduled API-key owner-route comment, a mid-session usage-credit gate owner-route comment, a false monthly-spend-limit owner-route comment, a billing-identity / limit-attribution owner-route comment, a usage-state / per-model weekly-limit owner-route comment, a usage-credit limit-enforcement / premature-completion owner-route comment, a session-wide premium-model billing-scope owner-route comment, a skill-loading / usage-window burn owner-route comment, an excessive Max-plan quota-burn owner-route comment, a usage-meter / limit-enforcement mismatch owner-route comment, a reset-window usage-limit mismatch owner-route comment, a no-use usage-window drain owner-route comment, a zero-meter weekly-limit lockout owner-route comment, a rapid weekly-limit spike owner-route comment, a zero-session-meter forced-credit-workaround owner-route comment, a bad-args 1000-agent fanout / cost-control owner-route comment, a cache-read metering / usage-meter owner-route comment, a cloud-review confirmation / cost-estimate owner-route comment, a workflow retry / duplicate usage owner-route comment, a subagent 1M-context / extra-usage entitlement owner-route comment, a task-scoped model-leakage / credit-exhaustion owner-route comment, a Max/Fable usage-credit gate owner-route comment, a Fable session-meter / usage-credit overflow owner-route comment, a Fable usage attribution / burn-rate owner-route comment, a Max entitlement / rate-limit owner-route comment, an org-seat entitlement / credit-prompt owner-route comment, and a central billing-pattern impact update for exact-match #80055 direct-credit-purchase failure, and the direct #80055 exact-match comment refreshed in place. These route/context records are useful because they map the same failure family across checkout, payment confirmation, support routing, credit-ledger provisioning, usage-meter reconciliation, session-limit accounting, paid-tier allocation, billing-credential resolution, public support visibility, local-cache state, per-model limit-bucket visibility, limit-enforcement / workflow-completion semantics, usage-meter enforcement mismatch, reset-window enforcement mismatch, no-use usage-window attribution, zero-meter weekly-limit enforcement, rapid weekly-limit spike enforcement, zero-session-meter forced-credit-workaround enforcement, bad-args agent-fanout cost-control, cache-read metering weight, and cloud-review billing-consent / documented confirmation-estimate boundaries, session-limit retry / duplicate paid-work attribution, and subagent context-tier / extra-usage entitlement boundaries, task-scoped model/credit-scope boundaries, Max/Fable usage-credit entitlement boundaries, Fable session-meter / usage-credit overflow boundaries, Fable usage attribution / burn-rate boundaries, Max entitlement / rate-limit boundaries, org-seat entitlement boundaries, and direct-credit-purchase impact boundaries while excluding private payment artifacts.

  • Central billing-pattern thread updated in place at 16:26 JST with today's #80055 public impact update, where the reporter says failed direct Anthropic credit purchasing forced a third-party workaround and capability loss, while retaining the request for a human Billing Platform / payment-operations reconciliation across checkout/payment confirmation, purchase commit, manual purchase versus auto-reload divergence, discount/prepaid-credit eligibility, credit-ledger posting, and support-case routing: anthropics/claude-code#55982 comment / #80055 impact update
  • Central billing-pattern thread updated in place at 15:47 JST with the latest #76427 Fable session-meter / usage-credit overflow and #67506 Fable usage attribution / burn-rate route evidence, while retaining the request for a human Billing Platform / payment-operations reconciliation across checkout/payment confirmation, purchase commit, discount eligibility, account controls, usage-ledger provisioning, and credit/cap restoration: anthropics/claude-code#55982 comment
  • Central billing-pattern thread updated again at 01:56 JST with the newly published public disappeared-credits and billing-transparency corroboration links, while retaining verified CrazyRouter/OpenClaw independent-guide corroboration plus #81116 session-limit / zero-token usage-meter and #72552 silent cross-account billing route evidence: anthropics/claude-code#55982 comment
  • Public X reply posted at 04:23 JST under the official Claude launch thread, tagging Anthropic with the sanitized manual-prepay-fails / auto-topups-work / no-billing-owner pattern and linking the public evidence hub: x.com/Coolak777/status/2081460261844828301
  • Facebook Claude Community billing/support-thread comment posted at 04:50 JST on a directly relevant public community thread, linking the sanitized hub and summarizing the manual prepaid-credit failure / auto-topups-work / no-billing-owner pattern: facebook.com/groups/claudecommunity/posts/1062395349634463/
  • Public Anthropic-owned subscription/extra-usage billing-mode reports added at 05:05 JST, covering claims that Max/Pro subscription usage or available plan quota can still drain Extra Usage or purchased credits despite subscription login or low plan usage: anthropics/claude-code#68501 and anthropics/claude-code#64613
  • Public Anthropic-owned cost-control route comments added at 05:26-05:38 JST, covering visible spend-ledger / overage enforcement disagreement, wrong monthly-spend-limit remediation for a different limiter, quota-consuming workflow interruption with no durable deliverable, ambiguous limiter errors that can leave background-agent work non-recoverable, and recursive agent fan-out without default spend guardrails: #39678 comment, #75730 comment, #79958 comment, #74006 comment, and #72732 comment
  • Public Anthropic-owned paid-tier allocation route comment added at 05:48 JST, covering a Max 5x to Max 20x upgrade allegedly not reflected in weekly limits and asking for Billing Platform / entitlement reconciliation: #79773 comment
  • Public Anthropic-owned phantom-usage / idle quota-drain / usage-leak route comments added at 11:57-12:06 JST, covering reports that paid Max usage can climb while idle or without visible activity logs, and asking for Billing Platform / cost-control / usage-accounting / agent-runtime owner reconciliation across server-side session lifecycle, quota debits, auto-reload triggers, phantom or orphaned work, and credit/cap restoration policy: #80199 comment and #80705 comment
  • Public Anthropic-owned billing-source / scheduled API-key route comment added at 12:17 JST, covering a Max subscriber silently routed from subscription billing to API billing for interactive and scheduled sessions, and asking for Billing Platform / entitlement / cost-control owner reconciliation across credential precedence, session billing source, scheduled execution, credit-ledger debits, subscription entitlement, and support routing: #78491 comment
  • Public Anthropic-owned mid-session usage-credit gate route comment added at 12:26 JST, covering Fable 5 becoming unusable mid-session with a usage-credit-required gate despite expected remaining access, and asking for Billing Platform / entitlement / usage-credit ledger owner reconciliation across model-access gates, subscription allowance, credit-ledger balance, manual purchase commit state, auto-reload state, session/runtime billing source, and support routing: #78614 comment
  • Public Anthropic-owned false monthly-spend-limit route comment added at 12:31 JST, covering a newly purchased subscription account blocked almost immediately by a monthly-spend-limit message after low reported usage, and asking for Billing Platform / entitlement / spend-limit owner reconciliation across subscription activation, spend-limit initialization, usage counters, credit-ledger balance, manual purchase commit state, auto-reload state, session/runtime billing source, and support routing: #79630 comment
  • Public Anthropic-owned billing-identity / limit-attribution route comment added at 12:36 JST, covering requests to show effective account/org/billing source and attribute limit messages to the right identity, and asking for Billing Platform / identity / entitlement / support-tooling owner reconciliation across credential precedence, account/org binding, session billing source, limit-message attribution, support-visible plan state, credit-ledger balance, manual purchase commit state, and auto-reload state: #77993 comment
  • Public Anthropic-owned usage-state / per-model weekly-limit route comment added at 12:45 JST, covering a dropped server-provided Fable weekly-scoped limit, contradictory local account flags, and another Claude surface showing the same limit correctly; asks for Billing Platform / usage-ledger / entitlement / Claude Code cost-surface owner reconciliation across server payloads, local cache, subscription flags, per-model limit buckets, reset windows, and support-visible state: #78507 comment
  • Public Anthropic-owned usage-credit limit-enforcement / premature-completion route comment added at 12:51 JST, covering a security scan reportedly exceeding configured usage-credit limits and reporting completion before usable work was finished; asks for Billing Platform / cost-control / usage-accounting / agent-runtime owner reconciliation across limit enforcement, workflow completion status, retry semantics, and credit/cap restoration policy: #80983 comment
  • Public Anthropic-owned session-wide premium-model billing-scope route comment added at 13:01 JST, covering Fable model selection reportedly applying to an entire session and billing routine turns at the premium rate without a clear confirmation boundary; asks for Billing Platform / cost-control / usage-accounting / Claude Code model-routing owner reconciliation across session-wide model changes, per-task routing, premium-rate confirmation UX, usage-ledger debits, and credit-review policy: #79478 comment
  • Public Anthropic-owned skill-loading / usage-window burn route comment added at 13:05 JST, covering a simple usage-limit/pricing question reportedly loading a very large bundled skill body and consuming disproportionate plan-window usage; asks for Billing Platform / cost-control / usage-accounting / Claude Code skill-loading owner reconciliation across skill trigger policy, deferred-loading behavior, usage-window accounting, and credit/usage-review policy: #79457 comment
  • Public Anthropic-owned excessive Max-plan quota-burn route comment added at 13:09 JST, covering Fable/Opus sessions reportedly consuming Max-plan quota much faster than the visible workload would suggest; asks for Billing Platform / cost-control / usage-accounting / Claude Code model-runtime owner reconciliation across model selection, session lifecycle, tool/subagent activity, cache behavior, quota debits, visible usage meters, and credit/usage-review policy: #78702 comment
  • Public Anthropic-owned usage-meter / limit-enforcement mismatch route comment added at 13:19 JST, covering a Max 5x report where Claude Code blocked input with Usage limit reached after roughly 35% weekly usage while Claude.ai web still worked on the same account; asks for Billing Platform / usage-ledger / entitlement / Claude Code limit-enforcement owner reconciliation across visible usage meters, enforcement gates, account/session billing source, credit-ledger state, and support-visible billing state: #78617 comment
  • Public Anthropic-owned reset-window usage-limit mismatch route comment added at 13:28 JST, covering a report where Claude Code said the user had hit a usage limit immediately after a session reset while the web surface showed a different state; asks for Billing Platform / usage-ledger / entitlement / Claude Code limit-enforcement owner reconciliation across reset-window state, web-vs-CLI usage state, local cache, enforcement gates, billing identity, credit-ledger state, and support-visible billing state: #78402 comment
  • Public Anthropic-owned no-use usage-window drain route comment added at 13:32 JST, covering a report where the 5-hour Claude Code usage window was reportedly consumed despite no same-day Claude Code use; asks for Billing Platform / usage-accounting / Claude Code runtime owner reconciliation across visible usage meters, server-side session lifecycle, account/session billing source, quota debits, credit-ledger state, enforcement gates, and support-visible billing state: #78087 comment
  • Public Anthropic-owned zero-meter weekly-limit lockout route comment added at 13:43 JST, covering a Max 20x report where Claude Code/Cowork enforced a weekly-limit block while visible weekly meters showed 0% and claude.ai web still worked on the same account; asks for Billing Platform / usage-ledger / entitlement / Claude Code limit-enforcement owner reconciliation across visible usage meters, hidden Claude Code-scoped limit buckets, reset-window state, account/session billing source, subscription entitlement, credit-ledger state, and support-visible billing state: #78982 comment
  • Public Anthropic-owned rapid weekly-limit spike route comment added at 13:53 JST, covering a report where weekly usage reportedly jumped from 98% to 100% faster than user activity would explain; asks for Billing Platform / usage-ledger / entitlement / Claude Code limit-enforcement owner reconciliation across visible weekly meters, hidden Claude Code-scoped counters, reset-window state, account/session billing identity, credit-ledger state, and support-visible state: #81245 comment
  • Public Anthropic-owned zero-session-meter / forced-credit-workaround route comment added at 14:04 JST, covering a Max 20x report where Claude Code enforced a usage-limit block while visible 5-hour session usage showed 0%, and the reporter said they had to purchase additional usage credits to work around the block; asks for Billing Platform / usage-ledger / entitlement / Claude Code limit-enforcement owner reconciliation across visible session meters, enforcement gates, subscription entitlement, account/session billing source, usage-credit ledger state, and refund/credit-restoration policy: #78352 comment
  • Public Anthropic-owned bad-args agent-fanout / cost-control route comment added at 14:16 JST, covering a report where a malformed workflow input triggered 1000 agent calls, burned 13M tokens, and hid an 800/1000 failure ratio behind a completion-shaped summary; asks for Billing Platform / cost-control / agent-runtime owner reconciliation across bad-args validation, agent fanout limits, non-retryable session-limit handling, token/credit debits, workflow completion semantics, and credit/cap restoration policy: #81351 comment
  • Public Anthropic-owned cache-read metering / usage-meter route comment added at 14:25 JST, covering a Max 20x report where transcript-level accounting after deduplication did not explain a 53% weekly quota drain unless cache-read or equivalent cached-context tokens were metered at a higher-than-expected weight; asks for Billing Platform / usage-ledger / cost-control owner reconciliation across per-request token categories, Max weekly and 5-hour limit weights, visible meters, hidden buckets, usage-credit debits, support-visible state, and credit/cap restoration policy: #81234 comment
  • Public Anthropic-owned cloud-review confirmation / cost-estimate route comment added at 14:41 JST, covering a report where a billed Claude Code cloud review launched from the CLI without the documented confirmation dialog, review scope, remaining free-run state, or estimated cost; asks for Billing Platform / cost-control / Claude Code cloud-review owner reconciliation across documented confirmation requirements, alias-specific CLI behavior, scope/cost estimate display, usage-credit debit timing, final charge attribution, support-visible state, and refund/credit-restoration policy: #77222 comment
  • Public Anthropic-owned workflow retry / duplicate usage route comment added at 14:51 JST, covering a report where a session-limit error was treated as retryable and subagents restarted cold from blank context, repeating completed work with no additional usable output; asks for Billing Platform / usage-accounting / cost-control / agent-runtime reconciliation across retry classification, duplicate-attempt attribution, repeated prompt/file-read/cache/tool tokens, artifact reuse, workflow preflight agent/model/cost confirmation, and refund/credit-restoration policy: #80253 comment
  • Public Anthropic-owned subagent 1M-context / extra-usage entitlement route comment added at 15:02 JST, covering a Max 20x report where Sonnet subagents, skills, and background agents appear to inherit a parent 1M-context request path and hit extra-usage / usage-credit gates even when the intended child work was standard 200K Sonnet; asks for Billing Platform / entitlement / cost-control / agent-runtime reconciliation across child context derivation, request-tier inheritance, failed admission-gate ledger treatment, extra-usage preflight, support-visible resolved model/context state, and credit/cap restoration policy: #68727 comment
  • Public Anthropic-owned task-scoped model-leakage / credit-exhaustion route comment added at 15:10 JST, covering a report where a Fable model request intended for one task allegedly became session-wide, drained usage credits, and hit an out-of-credits gate during compaction; asks for Billing Platform / cost-control / usage-accounting / model-routing reconciliation across per-task intent, persisted session model state, premium-rate confirmation, inherited turns/tools/subagents, ledger attribution, support-visible charge attribution, and credit/cap restoration policy: #79477 comment
  • Public Anthropic-owned Max/Fable usage-credit gate route comment added at 15:18 JST, covering a report where a paid Max account could not switch to Fable because Claude Code treated it as requiring usage credits; asks for Billing Platform / entitlement / model-routing reconciliation across Max-plan Fable entitlement state, client-side cached model access, standard included allowance vs extra-usage vs usage-credit-only path, account/org/billing identity, support-visible entitlement state, and credit/cap restoration policy: #80706 comment
  • Public Anthropic-owned Fable session-meter / usage-credit overflow route comment added at 15:30 JST, covering a report where Fable's dedicated 5-hour bar disappeared after reset while weekly Fable usage continued and overflow had already been billed to usage credits; asks for Billing Platform / usage-ledger / entitlement / Claude Code usage-UI reconciliation across Fable session-bucket lifecycle, generic session-bucket attribution, weekly Fable accounting, overflow debit trigger, hidden/stale meter state, support-visible ledger state, and credit/cap restoration policy: #76427 comment
  • Public Anthropic-owned Fable usage attribution / burn-rate route comment added at 15:47 JST, covering a long-running report where Fable-specific usage, delegated model work, background/session state, and usage-credit burn did not reconcile cleanly from user-visible counters; asks for Billing Platform / usage-ledger / entitlement / Claude Code runtime reconciliation across per-model attribution, background-agent lifecycle, Fable sub-cap accounting, usage-credit debit triggers, spend-limit gates, support-visible ledger state, and refund/credit-restoration policy: #67506 comment
  • Public Anthropic-owned Max entitlement / rate-limit route comment added at 15:52 JST, covering a long-running report where Claude Max subscription state and low visible usage still produced API rate-limit errors, with thread context discussing extra-usage/API-credit workarounds; asks for Billing Platform / entitlement / usage-ledger / rate-limit reconciliation across subscription recognition, server-side limiter state, visible usage counters, extra-usage/API-credit debit triggers, VS Code/CLI credential scope, account/org billing identity, support-visible state, and refund/credit-restoration policy: #29579 comment
  • Public Anthropic-owned org-seat entitlement / credit-prompt route comment added at 16:13 JST, covering a Team seat report where synced local entitlement state says hasAvailableSubscription: false and routes Claude Code toward credits / pay-as-you-go despite subscription entitlement; asks for Billing Platform / entitlement / auth-state reconciliation across org-seat vs personal subscription semantics, subscriptionType, model-access cache, CLI-vs-desktop entitlement divergence, credit-prompt gating, credential precedence, support-visible billing identity, ledger attribution, and refund/credit-restoration policy: #79671 comment
  • Independent Claude Code pricing-guide caveat verified at 04:10 JST: Dmytro Klymentiev's public guide now includes a billing-reliability caveat checked July 26, links Anthropic-owned tracker issues #73450, #56895, and #55982, says this is not a general outage, and advises users not to leave usage-credit top-ups until they are already blocked: klymentiev.com/blog/claude-code-pricing
  • Public Reddit billing/support-loop corroboration added at 02:13 JST, covering repeated emailed receipts absent from the Anthropic billing page, paid Max invoice state reverting to Free access, double-charged subscriptions with AI-support/refund friction, and unauthorized-upgrade/refund/account-disable claims; private identifiers from those threads are intentionally omitted here: random/repeated charge report, paid-Max downgraded-to-Free report, double-charged subscription report, and refund/account-disable support-loop report
  • Public Reddit usage-credit controls corroboration added at 02:26 JST, covering claims that promotional or free usage-credit claiming can interact with paid usage billing, spend-limit defaults, and unclear balance semantics; this is billing-control-surface context, not proof of this customer's private state: usage-credit controls warning
  • Public Reddit payment-initiation / credit-ledger corroboration added at 02:40 JST, covering API-credit purchase verification pages showing USD 0, top-up attempts charging $0 while balance remains zero, card declines where the bank reportedly saw no Anthropic attempts, paid extra-usage purchases not unlocking usage, usage-credit balances zeroing while $0 spent, and promotional credits disappearing after auto-billing hiccups; these are public pattern reports only: API-credit top-up $0 report, API-credit verification USD 0 report, card-decline / no-bank-attempt report, paid-extra-usage-not-unlocking report, usage-credit zeroed report, and promotional-credit disappearance report
  • Public Reddit billing-route / support-loop corroboration added at 02:54 JST, covering locked-credit/open-invoice payment failure with CVC-recollection checkout errors and Fin auto-closures, multi-card and wallet payment failures where an alternate higher-tier path reportedly succeeded, card/payment setup failures where banks reportedly saw no issue, gift subscriptions reverting to Free, extra-usage runtime errors despite visible usage/credit state, and hidden extra-usage billing-route reports; these are public pattern reports only: locked-credit invoice / CVC checkout report, multi-card wallet payment-failure workaround report, banks-see-no-issue payment setup report, gift subscription reverts-to-Free report, extra-usage runtime error despite visible state report, and hidden extra-usage billing-route report
  • Public Reddit and GitHub billing-state corroboration added at 03:10 JST, covering repeated upgrade payment failures after recurring-payment setup approval, Pro subscription payment-failure loops with trapped account credit and Fin handoff cycling, failed-payment email floods after a paid subscription and no reported API use, paid-invoice/past-due access mismatches despite paid invoices or additional Console credit, unexpected annual-plan charge/support-bot lockout reports, and official refund-link / redeemed-credit state transitions where credit or subscription payment allegedly did not reflect; these are public pattern reports only and private identifiers from source material are intentionally omitted here: upgrade payment failure after setup approval report, Pro payment loop / trapped credit report, failed-payment notification flood report, paid-invoice / past-due mismatch report, unexpected annual-plan charge support-loop report, and refund-link / credit-state transition report
  • Public Reddit large-billing-error / refund-routing corroboration added at 03:26 JST, covering a reported Anthropic-acknowledged large billing-system error with repeated bank charge attempts and automated escalation routing, plus a refund-window support-bot report that was later escalated to a person and refunded; these are public pattern reports only and private identifiers from source material are intentionally omitted here: large billing-system error / escalation-routing report and refund-window support-bot escalation report
  • Public GitHub known-billing-system-error / manual-intervention corroboration added at 03:36 JST, covering a public Anthropic-owned issue that reports support-bot acknowledgement of a known July 2026 billing-system error affecting multiple customers and a human-support statement that missing credits required investigation/manual intervention; this is public pattern evidence only and private identifiers from source material are intentionally omitted here: #77097 known billing-system error / manual-intervention issue
  • Public GitHub mid-cycle downgrade / unauthorized plan-mutation corroboration added at 03:44 JST, covering public Anthropic-owned issues that report paid accounts being downgraded or plan state mutating without intended user action, double-charge/refund friction, generated Gift Max invoices, and support-routing gaps; this is public pattern evidence only and private identifiers from source material are intentionally omitted here: #64557 mid-cycle downgrade / double-charge issue and #51168 unauthorized plan-mutation issue
  • Public GitHub free-plan downgrade / gift-cancellation / unauthorized gift-charge corroboration added at 03:49 JST, covering public Anthropic-owned issues that report paid Max accounts reverting to Free, gift subscriptions auto-canceling with broken redemption/credit links, unauthorized gift-subscription charges, disappeared support conversations, and extended support silence; this is public pattern evidence only and private identifiers from source material are intentionally omitted here: #41581 paid Max downgrade thread, #44163 gift cancellation / redemption failure issue, and #62461 unauthorized gift-charge / support-loss issue
  • Central billing-pattern thread updated in place at 04:10 JST with the verified Dmytro pricing-guide caveat plus the same #41581 / #44163 / #62461 paid-plan, gift-entitlement, and support-loss corroboration, while noting that public status still showed All Systems Operational at 03:59 JST and repeating the request for a human Billing Platform / payment-operations owner: anthropics/claude-code#55982 comment
  • Prepaid commit / zero-bank-auth thread refreshed at 22:37 JST: anthropics/claude-code#73450 comment
  • Banks-see-zero-attempts / support-routing thread refreshed at 22:40 JST: anthropics/claude-code#54055 comment
  • Fin auto-closes / organization-credit thread refreshed at 22:42 JST: anthropics/claude-code#80178 comment
  • Card-auth-succeeds / Anthropic-purchase-fails thread refreshed in place at 16:31 JST with today's #80055 public impact update and central anchor link: anthropics/claude-code#80055 comment / #80055 impact update / central anchor
  • Console Buy Credits path disabled / inconsistent billing-state thread refreshed at 22:51 JST: anthropics/claude-code#62644 comment
  • Payment-fails-across-cards / support-unresponsive thread refreshed at 22:57 JST: anthropics/claude-code#56281 comment
  • Balance-label / credit-ledger reconciliation thread added at 23:09 JST: anthropics/claude-code#67083 comment
  • Paid-plan usage-meter / reset-boundary / support-routing report linked at 23:36 JST: anthropics/claude-code#81366 comment
  • Session-limit / zero-token usage-meter mismatch linked at 00:44 JST: anthropics/claude-code#81116 comment
  • Silent cross-account billing / credential-resolution route linked at 00:57 JST: anthropics/claude-code#72552 comment

Historical Same-Method Account Lockout Evidence

A closed public Anthropic-owned issue from 2026-05-24 reports the same payment method being accepted on one Anthropic account while rejected on another, followed by Fin handoff to human support and no visible resolution before stale closure. This is public historical pattern context for account-level billing, risk, and support-routing divergence, not proof of private payment artifacts.

Same-Card Manual Payment / Paid-Entitlement Evidence

Additional open Anthropic-owned issues show the same failure family from two angles: card acceptance differs by billing path, and paid billing records can fail to map to usable Claude access. These links are public pattern context only, with private identifiers intentionally omitted.

  • Same card reportedly used for the regular monthly bill, but a manual payment attempt was declined and later led to an account-access disabled state: anthropics/claude-code#72178
  • Paid Max invoice/account-dashboard mismatch where support reportedly acknowledged backend synchronization without a public resolution timeline: anthropics/claude-code#71821
  • Paid Max invoice followed by a Free-plan state and support ticket submission, with private invoice details omitted here: anthropics/claude-code#68257

Post-Payment Access-Disabled Evidence

An open Anthropic-owned issue with cost, auth, and oncall labels reports payment/subscription activity followed by disabled Claude access, refund/support friction, and a long public thread of similar user reports. This is public pattern context only; this matrix does not copy user emails, account identifiers, screenshots, appeal details, or invoice/payment identifiers from the thread.

  • Payment or renewal followed by disabled Claude/organization access and unresolved support-routing reports: anthropics/claude-code#5088
  • Paid Max account routed into account-creation/onboarding access failure after payment, while support reportedly auto-closed; author later reported an account-creation workaround, so this is historical account-state/support-routing evidence only and private receipt/email/account details are not copied here: anthropics/claude-code#51294
Fails

Manual Prepay

Manual usage-credit checkout fails or loses discount state after customer-side steps.

Works

Auto-Reload

Automatic usage-credit top-ups continue charging through the saved billing setup.

Missing

Owner Review

No confirmed human Billing Platform / payment-operations owner or written reconciliation yet.

Reconciliation Matrix

Failure Lane What It Shows Anthropic Systems To Inspect Representative Public Context Resolution Test
Manual prepaid / bulk credit purchase fails while auto-reload works The account can be charged automatically, but manual prepay fails or loses the advertised discount. This points away from a simple insufficient-funds diagnosis. Checkout session, payment-confirmation state, purchase commit, invoice/order state, discount eligibility, risk/radar decisioning, usage-credit ledger. #73450, #80178, #80055, #81345, #81345 comment, #72178, #79458, #61081, #72411, #45919, #45919 comment, #25745, #57122, #54752, #55266, #75299, #75299 comment, #75640, #66365, #54976, #73982, #54204, #58832, #60232, Reddit API-credit $0 top-up report, Reddit verification USD 0 report, Reddit card-decline / no-bank-attempt report, Reddit multi-card wallet payment-failure workaround report, Reddit banks-see-no-issue payment setup report, Reddit upgrade payment failure after setup approval report, Reddit Pro payment loop / trapped credit report, #54055, #54055 comment Manual prepay works at the advertised terms, or Anthropic applies an equivalent adjustment and explains the failed manual state.
Paid invoice, charge, or credit purchase does not become usable balance Billing documents, charges, or prepaid annual credit can exist while product entitlement, balance, refund, or credit-ledger state remains wrong. Invoice finalization, receipt/order mapping, credit-ledger posting, subscription status propagation, entitlement propagation, auth-state sync, annual-credit consumption, refund amount calculation, cancellation state, account/workspace mapping, delayed reversal jobs. #56895, #41581, #44163, #62461, #64557, #51168, #64480, #64480 comment, #66558, #71821, #68257, #5088, #51294, #56897, #75475, #60923, #60923 comment, #62991, #48231, #46378, #41499, #45335, #52396, #43257, #55266, #58101, #61339, #48780, #68429, #31537, #78905, #80722, #70714, #72027, #72027 comment, #69786, #69786 comment, #77444, #77444 comment, #79566, #45497, #23674, #77097, #31012, #10832, Reddit missing-credits report, Reddit disappeared-credits report, Reddit billing-transparency report, Reddit paid-extra-usage-not-unlocking report, Reddit locked-credit invoice / CVC checkout report, Reddit gift subscription reverts-to-Free report, Reddit paid-invoice / past-due mismatch report, Reddit unexpected annual-plan charge support-loop report, Reddit large billing-system error / escalation-routing report, Reddit refund-window support-bot escalation report, #45662, #52138, #44720, #17694, #64555, #44198, #56338, #65514, #65514 comment Paid funds map to usable Claude balance or entitlement, with a written explanation for any reversal, delay, or workspace mismatch.
Visible credits are rejected by API/runtime entitlement Account or workspace balance can appear funded while API/runtime requests still fail as if no credits exist. Credit-ledger propagation, API entitlement cache, workspace/org mapping, auth-state sync, credit-grant availability, runtime billing-gate error mapping. #54839, #54839 comment, Reddit extra-usage runtime error despite visible state report Visible funded balance maps to usable API/runtime access, or Anthropic explains the exact workspace, grant, limit, or propagation reason for rejection.
Subscription entitlement and visible usage disagree with rate-limit gates Claude Max subscription state and low visible usage can coexist with API rate-limit errors, while extra-usage or API-credit workarounds muddy whether the blocker is entitlement, rate-limit state, or credit-ledger state. Subscription entitlement service, rate-limit evaluator, visible usage counters, usage-ledger attribution, extra-usage/API-credit admission gate, VS Code/CLI credential scope, account/org billing identity, support-visible plan and limiter state. #29579, #29579 comment Anthropic identifies the exact server-side limiter and billing source being enforced, reconciles it against visible usage and subscription entitlement, and restores credits or caps where an internal mismatch drove paid workarounds.
Spend-limit, extra-usage, and usage-credit messaging conflict Users are pushed toward credit purchases when the real blocker may be a monthly limit, spend cap, crossover setting, included-plan quota state, session-limit counter, usage-window attribution state, hidden fanout burn, zero-session-meter enforcement, hidden weekly-limit bucket, or cache-read metering weight not reflected in visible usage totals. Spend-limit evaluator, extra-usage settings, usage-credit balance display, CLI error mapping, plan quota service, model routing, monthly limit reset state, session-limit service, usage-window accounting, bad-args validation, agent fanout limits, cache-read metering weights, hidden weekly-limit buckets, subagent attribution, refund/credit-restoration policy. #74784, #77819, #79630, #79630 comment, #81309, #81366, #81366 comment, #81116, #81116 comment, #76427, #76427 comment, #67506, #67506 comment, #78507, #78507 comment, #80199, #80199 comment, #80705, #80705 comment, #81359, #81351, #81351 comment, #81245, #81245 comment, #81234, #81234 comment, #79477, #79478, #79478 comment, #79457, #79457 comment, #78702, #78702 comment, #78617, #78617 comment, #78402, #78402 comment, #78087, #78087 comment, #78982, #78982 comment, #78186, #72131, #69430, #32544, #39678, #23579, #74851, #75730, #74006, #74006 orphaned-work comment, #74006 paid-credits comment, #68285, #78231, #72732, #72861, #79958, #80983, #80983 comment, #65514, #65514 comment, #45925, #76006, #79167, #67412, #78772, #73795, #75757, Reddit usage-credit controls warning, Reddit usage-credit zeroed report, Reddit promotional-credit disappearance report, Reddit extra-usage runtime error despite visible state report, Reddit hidden extra-usage billing-route report UI and CLI identify the actual blocker and do not direct users to buy credits when credits cannot solve the account state.
Chargeable cloud review launches without documented confirmation A billed cloud workflow can debit usage credits before the user sees the documented confirmation dialog, review scope, remaining free-run state, or cost estimate. Cloud-review launch path, alias routing, confirmation dialog gating, review scope estimator, free-run accounting, usage-credit ledger debit timing, support-visible charge attribution, refund/credit-restoration policy. #77222, #77222 comment Chargeable review paths show and require acceptance of the documented confirmation/estimate boundary before launch, or Anthropic restores credits when the boundary was skipped.
Session-limit retry duplicates paid workflow usage A known session-limit error can be treated as retryable, causing subagents to restart from blank context, repeat completed work, and debit plan or credit usage without additional usable output. Agent-runtime retry classifier, workflow state preservation, artifact reuse, subagent attribution, model allocation/preflight estimator, usage-meter and credit-ledger debits, support-visible charge attribution, refund/credit-restoration policy. #80253, #80253 comment Session-limit and quota errors pause instead of cold-retrying until reset or explicit user confirmation; duplicate attempts are attributed and restored when platform retry produced no additional output.
Subagent context-tier inheritance crosses into extra usage Plain Sonnet subagents, skills, or background agents can inherit a parent 1M-context request path and hit extra-usage / usage-credit gates even when the intended child work was standard 200K Sonnet. Child-agent request construction, context-tier derivation, model entitlement service, extra-usage admission gate, usage-credit ledger, failed-attempt attribution, support-visible resolved model/context state, credit/cap restoration policy. #68727, #68727 comment Child agents default to standard context unless explicitly configured otherwise; any extra-usage crossover is preflighted and failed admission-gate attempts are not treated as customer-consented paid work.
Zero visible session usage still triggers credit workaround Claude Code can enforce a usage-limit block while the visible 5-hour session meter shows 0%, then push a Max subscriber toward buying additional usage credits. Visible session meters, enforcement gates, subscription entitlement, account/session billing source, usage-credit ledger state, support-visible state, refund/credit-restoration policy. #78352, #78352 comment Anthropic explains the 0%-meter limit decision, restores any wrongly consumed credit value, and prevents paid-credit prompts from masking true plan/session entitlement state.
Plan or model entitlement is not recognized and routes to usage credits A paid plan, org seat, or available quota can be visible elsewhere while Claude Code or model routing demands usage credits instead. Plan entitlement service, org-seat entitlement semantics, hasAvailableSubscription / subscriptionType flags, model-access cache, Claude Code account binding, workspace/org resolution, quota meter routing. #80836, #79516, #80889, #80334, #80813, #81307, #81350, #79597, #80749, #79386, #79341, #79412, #79351, #79441, #78610, #78620, #78611, #78634, #78614, #78614 comment, #79548, #79548 comment, #80706, #79360, #80382, #73646, #77219, #70609, #62063, #39841, #69154, #79337, #74051, #80484, #80484 comment, #79413, #79413 comment, #80409, #80737, #80737 comment, #79576, #79576 comment, #70721, #74379, #74388, #79410, #81318, #34296, #80043, #79332, #79460, #80750, #79671, #79671 comment, #79687 Claude surfaces agree on plan, model, quota, and usage-credit state, with no hidden reroute to credits when plan quota should apply.
Credential, account, or support-visible state silently changes billing source Users can be logged into or paying for one Claude plan while runtime, stored credential, API organization, or support lookup state routes them to a different billing source. Credential precedence, environment-token handling, stored API-key state, child-process credential inheritance, Claude account binding, Console/API organization mapping, org-seat entitlement state, hasAvailableSubscription / subscriptionType flags, model-access cache, Fin account lookup, billing-source display. #79602, #78491, #78491 comment, #80713, #80713 comment, #62338, #62338 comment, #77617, #60901, #74217, #77993, #77993 comment, #72552, #72552 comment, #77376, #23315, #79427, #80973, #80973 comment, #67083, #67083 comment, #79671, #79671 comment, #79687, Reddit hidden extra-usage billing-route report Claude surfaces and support tooling identify the same account, plan, organization, and billing source before routing users to credits or metered API spend.
Support routing and dead escalation paths block reconciliation Fin or support can acknowledge the need for specialized billing review but still close, deflect, route users through non-working addresses, or offer automated refund/cancellation paths that do not expose the full prepaid-credit ledger. Fin handoff rules, support case ownership, payment-ops queue routing, failed-payment sender configuration, refund/cancellation automation, prepaid annual-credit visibility, escalation SLA, case closure automation. #79598, #41581, #44163, #62461, #68773, #62533, #62533 comment, #81078, #56281, #54055, #54055 comment, #81345, #81345 comment, #60923, #60923 comment, #62010, #5088, #79566, #45497, #77097, #64557, #51168, Reddit disappeared-credits report, Reddit billing-transparency report, #80973, #80973 comment, #68429, #68429 comment, #78577, Reddit locked-credit invoice / Fin auto-closure report, Reddit banks-see-no-issue support-bot report, Reddit multi-card wallet support-silence report, Reddit Pro payment loop / trapped credit report, Reddit failed-payment notification flood report, Reddit unexpected annual-plan charge support-loop report, Reddit large billing-system error / escalation-routing report, Reddit refund-window support-bot escalation report A named human owner or team responds with payment-ops findings, not another generic card-decline article or dead reply address.

Minimum Owner Query

  • Compare the same account's manual usage-credit checkout records against automatic top-up records.
  • Confirm whether any failed manual purchase consumed discount eligibility or created hidden payment/order state.
  • Trace the checkout through payment confirmation, purchase commit, invoice/order creation, and credit-ledger provisioning.
  • Confirm whether CLI or UI limit messages route the user to credits when spend-limit or plan quota state is the actual blocker.
  • Compare server-provided usage-limit payloads, local cache, displayed account flags, and Claude surface usage views for the same account before treating a missing limit bar as only a client rendering bug.
  • Reconcile configured usage-credit limits, actual credit/quota debits, completion status, and retry requirements before deciding whether paid work should consume additional credits.
  • Reconcile session-wide model changes, per-task routing intent, premium-rate confirmation UX, usage-ledger debits, and credit-review policy before treating high-cost model usage as consented quota burn.
  • Reconcile skill trigger policy, deferred-loading behavior, usage-window accounting, and credit/usage-review policy before treating internal tool-loading overhead as customer-consented quota burn.
  • Reconcile model selection, session lifecycle, tool/subagent activity, cache behavior, quota debits, visible usage meters, and credit/usage-review policy before treating rapid quota drain as expected customer usage.
  • Reconcile visible usage meters, enforcement gates, account/session billing source, credit-ledger state, and support-visible billing state before treating a usage-limit block as true customer exhaustion.
  • Reconcile visible 5-hour session meters, enforcement gates, subscription entitlement, account/session billing source, usage-credit ledger state, and refund/credit-restoration policy before treating a zero-session-meter usage-limit block as true customer exhaustion or a customer-consented credit purchase.
  • Reconcile reset-window state, web-vs-CLI usage state, local cache, enforcement gates, billing identity, credit-ledger state, and support-visible billing state before treating a post-reset usage-limit block as true exhaustion.
  • Reconcile visible usage meters, server-side session lifecycle, account/session billing source, quota debits, credit-ledger state, enforcement gates, and support-visible billing state before treating a no-use usage-window drain as valid customer usage.
  • Reconcile visible usage meters, hidden Claude Code-scoped limit buckets, reset-window state, account/session billing source, subscription entitlement, credit-ledger state, and support-visible billing state before treating a zero-meter weekly-limit lockout as true exhaustion.
  • Reconcile visible weekly meter changes, hidden Claude Code-scoped counters, reset-window state, account/session billing identity, credit-ledger state, and support-visible state before treating a rapid weekly-limit spike as true customer exhaustion.
  • Reconcile per-request fresh input, cache-read, cache-write, and output token categories against Max weekly and 5-hour limit weights before treating a transcript-deduped cache-heavy session as true weekly exhaustion.
  • Reconcile documented confirmation/estimate requirements, alias-specific CLI behavior, review scope/cost display, usage-credit debit timing, final charge attribution, and refund/credit-restoration policy before treating a no-confirmation cloud review charge as customer-consented spend.
  • Reconcile session-limit retry classification, duplicate-attempt attribution, repeated prompt/file-read/cache/tool tokens, artifact reuse, workflow preflight agent/model/cost confirmation, and refund/credit-restoration policy before treating cold-retry workflow usage as valid customer usage.
  • Reconcile child-agent context derivation, request-tier inheritance, failed admission-gate ledger treatment, extra-usage preflight, support-visible resolved model/context state, and credit/cap restoration policy before treating standard-context subagent work as customer-consented extra-usage.
  • Confirm whether environment tokens, stored API keys, or account/org binding can silently change the billing source away from the paid subscription.
  • Confirm whether Fin/support account lookup sees the same paid plan and organization state as Claude web, desktop, Console, and Claude Code.
  • Confirm the working support route for payment-operations review, because the failed-payment sender reply path bounced as address-not-found.
  • Compare public paid-state/support-loop reports involving repeated Anthropic receipts absent from the billing page, paid Max invoice state reverting to Free access, double-charged subscriptions, and refund/account-disable loops against the same internal invoice, entitlement, refund, and support-routing services implicated by this case.

Independent Buyer-Guidance Context

Independent Claude/Anthropic guide sites have started adding public buyer-guidance caveats that separate manual credit purchases from automatic reloads. CrazyRouter's payment-guide update, reverified live at 2026-07-27 00:38 JST, frames manual prepaid-credit failure while auto-reload still works as a distinct troubleshooting case; Dmytro Klymentiev's verified Claude Code pricing guide adds a billing-reliability caveat linked to Anthropic-owned tracker issues; and OpenClaw's verified guides tell readers to verify one-time purchase and auto-reload paths separately before depending on direct Anthropic Console billing.

Public Evidence Links

Privacy Boundary

This public matrix intentionally omits card endings, bank names, payment processor object IDs, invoice or receipt identifiers, raw browser/network logs, screenshots, payment URLs, one-time codes, private support-thread bodies, addresses, identity documents, document numbers, and agency report numbers.