Skip to content

Owner & Administrator Guide

This guide walks through every screen. Demo mode (--demo, or the "Demo Mode" Start-menu shortcut) is the safest way to learn the tool — it uses a fully anonymized synthetic organization.

Global controls (left sidebar)

  • Reporting period — This month / Last month / This week / Last week / This quarter / This year / Today / Custom range. Every page and every export follows this selection.
  • Sync now — incremental synchronization (see Sync & Audit).
  • Refresh view — rebuilds the current data snapshot.
  • The status bar always shows data freshness, the active scoring policy version, and a ⚠ stale-data warning when the last sync is older than the configured threshold.

1 · Executive Dashboard

KPIs (seats, active/inactive users, seat utilization, average & median Weighted Usage %, low-usage and exempt counts, chat messages, Claude Code sessions, tokens, cost), the daily usage trend, users-by-status, product adoption and daily cost. Hover any chart for exact values.

2 · User Usage

Searchable, sortable table of every account: role, membership status, group, Chat/Code/Other/Total usage %, active days, last activity, cost, usage status and exemption. Filters: text search, membership status, usage status, group.

Double-click a user for the drill-down profile: daily activity chart, raw metrics, and the full Weighted Usage Index breakdown (weight, effective target, actual, uncapped attainment, capped contribution, and whether each metric was applicable — with the reason when it wasn't).

3 · Product Analytics

Adoption per product (active users, top metrics, tokens, cost), the full metric list per product, token/cost by model, and skill adoption. Products newly returned by the API appear automatically.

4 · Low-Usage Alerts

  • Current evaluation — live classification of the selected period (critical / warning / no-data / grace / exempt) with reasons.
  • Evaluate current period now — persists alerts (respecting per-period dedupe and the suppression window) and reports what was created/skipped.
  • Alert records & history — every persisted alert with state (open / acknowledged / resolved), follow-up status, and comments. Select rows to Acknowledge, Resolve, set a follow-up status, add comments, or Export selected alerts (CSV).

Alerts are informational only. The application never changes accounts.

5 · Reports

  1. Choose the template: Executive Management Summary, Detailed User Usage Report, or Low-Usage & Exception Report.
  2. Choose the period and optional filters (usage status, group, "exclude exempt from totals").
  3. Tick the formats: PDF, XLSX, DOCX, PPTX, CSV, HTML, JSON.
  4. Refresh preview — the preview and the exports render from the same snapshot, so exported numbers match exactly what you see.
  5. Generate report — files land in the reports folder (Open reports folder) with professional names like Claude_Organization_Usage_Executive_2026-08.pdf.

Every report carries: organization name/logo, generation timestamp, period, timezone, applied filters, data sources, last-refreshed time, scoring-policy version, limitations, page numbers (PDF/DOCX) and the confidential footer. Report history lists every generation with status and files.

The Excel workbook contains the full worksheet set: Executive Summary, Users, Daily Trends, Product Usage, Claude Code, Tokens & Cost, Low Usage, Exemptions, Scoring Methodology, Data Sources, Audit Info.

6 · Settings

  • API Connections — select your Organization plan (Team / Enterprise / Console) so guidance and unavailability reasons match what Anthropic actually offers on your plan; add/replace/remove the Analytics and Admin keys (Credential Manager), test the connection, and review the capability matrix (which categories your plan serves, from which source, and why anything is unavailable), plus the connected organization identity. Team plan: both key fields stay empty by design — see the operating notes below.
  • Scoring — the Weighted Usage Index: basis days, period scaling, and per-metric enabled/weight/target/cap. Add custom metrics (product + metric key, or active_days). Saving creates a new policy version; all screens re-score immediately, history included.
  • Alert Thresholds — passing/warning thresholds, evaluation period, minimum active days, grace period, evaluation cadence, suppression window, weekend/holiday handling, which membership states and service accounts are evaluated, no-data alerting, and notification channels.
  • Alert Exemptions — the dedicated exemption register: exact email, wildcard (service-*@company.com), domain (external-domain.com) or group; reason, approver, start date, permanent or expiring, notes. Deactivate with confirmation; import/export the list as CSV. Exempt accounts stay visible everywhere and return to normal evaluation when the exemption expires. Every change is audited.
  • Email Delivery — SMTP (STARTTLS/SSL) or Microsoft 365 / Graph (client-credentials; secret in Credential Manager), recipients/CC, subject & body templates (Jinja2 variables listed in the UI), weekly and monthly low-usage summaries, and a test-email button.
  • Branding — organization name, report title, confidentiality label, primary color, logo (used across the UI and all reports).
  • Data & Backup — timezone (default Asia/Phnom_Penh), light/dark/system theme, logging level, privacy notice, retention (metrics / audit / API log / backups), encrypted backups toggle, Back up database now, and Restore from backup… (validated, with a .pre_restore safety copy).
  • Scheduling — automatic sync interval, revision lookback, initial/full history depth, stale-data threshold; scheduled report generation (weekly/monthly, template, formats, optional emailing of the files).

Always finish with Save all settings.

7 · Sync & Audit

  • Sync now / Full resynchronization / Cancel with live per-category progress. Full resync re-fetches the configured history window; all writes are idempotent upserts, so nothing is ever double-counted and a failed run never damages the last good dataset.
  • Sync history — every run with per-category outcomes (double-click for details), errors, and trigger (manual/scheduled).
  • Official export imports — choose a profile (members, claude.ai members analytics, chat daily, Claude Code daily, token/cost daily, claude.ai spend report, generic long-format), then pick or drag-and-drop a CSV. The members-analytics export has no date column — one row covers the whole window — so its reporting window is read from the file name (…20260802to20260831.csv) and every metric is stored on the window's end date; keep the name claude.ai gives the file. A preview shows the detected column mapping, sample rows and validation problems before anything is written. Results report imported values, rows skipped because official API data already covers them, rejected rows (stored with reasons — double-click a batch), and emails not found in the directory.
  • API requests — sanitized request telemetry (never credentials).
  • Audit trail — every configuration/exemption/policy change, sync, import, report and alert action.

Operating on the Claude Team plan

The Team plan has no Analytics or Admin API (verified against the official Anthropic documentation), so this application runs key-less there — the data arrives through the official exports instead of a sync:

  1. Once: Settings → API Connections → set Organization plan to Claude Team plan; keep a member list current with the Organization members export profile.
  2. Each reporting period (weekly or monthly):
  3. Export Claude Code analytics from claude.ai/analytics/claude-code (Export all users) and import it with the Claude Code daily usage export profile.
  4. Export the spend report (token usage and estimated spend per user and per model) from your organization's analytics settings and import it with the claude.ai spend report profile — rows are summed across models per user per day.

  5. Refresh the view; dashboards, the Weighted Usage Index, alerts and all report formats work identically on imported data. Metrics the exports do not contain (e.g. chat message counts) are shown as Unavailable / No Data and are excluded from the scoring denominator — consider disabling those metrics on the Scoring tab so weights concentrate on the data you actually receive. A browser-only shortcut: the Usage analyzer on this site runs the same index, the same "unavailable is never zero" rule and the same threshold bands over the files you drop into it, entirely on your device. Use it to settle on thresholds before configuring them here, or for a quick month-end read without opening the application.

  6. If your company also operates a Claude Console (API) organization, add its Admin API key: Claude Code analytics for that organization then syncs automatically, and imports remain for the claude.ai side.

  1. Daily/6-hourly: automatic incremental sync (Settings → Scheduling).
  2. Weekly: review Low-Usage Alerts, acknowledge and set follow-ups.
  3. Monthly: generate the Executive Summary (PDF+XLSX) and the Low-Usage report; review exemptions for expiry; back up the database.
  4. Quarterly: revisit scoring targets/weights with management — the policy version history keeps prior reports defensible.