Observability

Dashboards

The organization-scoped Vectoralix dashboard surfaces request logs and rolled-up metrics as a small set of widgets. This page walks through each widget, what it answers, the time windows it supports, and how to read its empty state.

Where the widgets live

Open the Vectoralix service dashboard at /app/{tenant}/dashboard for the organization you want to view. The metrics widgets sit alongside the existing subscription stats and the per-server quota widget, in the same scrollable column.

Status — partly roadmap

The widgets described below are the planned shape of the dashboard. The underlying pipeline — request logging and rollup aggregation — is wired up; the widgets that surface those numbers to organization members are still being delivered. Treat this page as the target reading; widget styling and time-window defaults may shift before launch. The admin-facing platform metrics are out of scope here.

Overview stats

A four-tile stats row at the top of the dashboard: requests in the last hour, last 24 hours, last 30 days, and the count of "active" MCP servers. "Active" means the server served at least one request inside the displayed window, so a server that has been idle for the period drops out of the count even if it is configured and running.

Requests-over-time chart

A line chart of requests over time, with a window toggle between 24 hours, 30 days, and 12 months. Each series corresponds to one MCP server in the organization, with an aggregated total on top. The chart's data source switches with the window — hourly rollups for 24h, daily for 30d, monthly for 12m — so longer windows render in constant time.

Top Tools table

A table of the ten most-invoked tools across the organization in the selected window, with request counts. Tools that share a name across MCP servers are listed per server, not merged — so two servers that both expose a "search_files" tool show up as separate rows.

Requests-per-Server table

A breakdown of request counts across every MCP server in the organization, in the same window as the rest of the dashboard. Use this to spot the busiest server when planning capacity, and to spot an unexpectedly idle server when debugging an integration.

Tenant scoping

  • Every widget is fenced to the currently selected organization. Switching tenants reloads all widgets with that tenant's data.
  • Non-members never reach this dashboard route — the panel's membership gate runs before the widgets render.
  • There is no way to view another organization's metrics from this dashboard, even for a member of multiple organizations — switch the tenant first.

Timezones on the viewer side

Chart axes label time in the viewer's local zone. The numbers themselves do not change across timezones — only the labels. A row that says "April 25, 14:00" for a viewer in UTC will say something else for a viewer in Pacific time, but it is the same row.

Empty state

A new organization with zero traffic sees zeros across the four overview tiles, a "No requests yet" caption on the chart, a flat baseline, and empty Top Tools / Requests-per-Server tables. This is intentional, not a bug — there is nothing to display until at least one MCP request has been served.

Dashboard vs landing-page stats

The marketing stats on the public Vectoralix landing page are platform-wide and counted from the admin-side rollups. The numbers on this dashboard are scoped to a single organization. Do not expect the two to match — they answer different questions and read different sources.

See also: Request Logs explains the underlying events these widgets count, Usage Metrics explains the rollup pipeline that feeds the longer windows, and the Plans & Billing page on the landing site documents the quota widget that sits next to these on the same dashboard.