Skip to content

AI Inventory

AI Inventory is a consolidated, read-only inventory of the AI that flows through Arbitex — the providers, models, MCP servers, and retrieval sources you’ve configured, joined with usage telemetry. Arbitex is an AI gateway, not an inline network appliance: it reports only what the control plane can actually observe, and it does not claim to discover AI usage that bypasses the gateway.

The AI Inventory panel showing summary tiles for total inventory items and per-kind counts, followed by a filterable table of providers, models, MCP servers, and retrieval sources with usage and last-seen columns.
Monitoring & Reporting → AI InventoryCitadel synthetic data on a non-production Arbitex demo instance. Never client data.
  1. Open Monitoring & Reporting → AI Inventory. Review the Sanctioned AI estate summary tiles — total items and a count per kind (providers, models, MCP servers, retrieval sources).

  2. Filter the table by Kind or Status (active, idle, configured, unused).

  3. Read the Usage (requests) column: models show a real measured count (including a real zero); other kinds show “not measured” — never a fabricated number — because only per-model usage telemetry exists today.

  4. Review Last seen, Status, Provider, and Scope for each row; export the table via the shared column-chooser/CSV export.

Terminal window
curl -X POST "https://admin.arbitex.ai/api/v1/admin/monitoring/ai-inventory" \
-H "Authorization: Bearer $ADMIN_TOKEN" \
-H "Content-Type: application/json" \
-d '{}'

Available on every plan.

The inventory composes existing provider/model/MCP/retrieval-source configuration with usage telemetry server-side. In an air-gapped deployment it reflects only what the local control plane has configured and recorded.