Observability # Overview Vectoralix records one row per MCP request the moment it is served, rolls those rows into hourly, daily, and monthly summaries on a background scheduler, and surfaces both live and rolled-up numbers on the organization dashboard. This group explains that pipeline from the reader's side: what a single request writes down, how long the raw rows live, how the rollups are built and when they lag, and what users actually see when they ask "is my server being used, and by whom." ## What Observability covers Observability is the answer to three questions every MCP server owner asks once their endpoint is live: who is calling it, how often, and how is it performing. Vectoralix answers all three from a single source of truth — one append-only row per request — and a small set of rollup tables that summarize that row stream into the windows a dashboard cares about. ## The pipeline at a glance Every request to /mcp/<serverUid> writes a row to request\_logs. Scheduled jobs read those rows and produce hourly, daily, and monthly summaries. The dashboard reads either the raw rows or the summaries, depending on the time window you ask for, so short-window numbers are second-fresh and long-window charts are constant-time to render. ``` MCP request → request_logs (raw, 30-day retention) ↓ MetricsAggregator (scheduled rollups) ↓ request_metrics_hourly / _daily / _monthly (kept indefinitely) ↓ Dashboard widgets ``` ## Pages in this group Page What it answers [Request Logs](https://vectoralix.com/docs/observability/request-logs) What gets recorded for every MCP request, what fields exist, how long rows live, and how to read them. [Usage Metrics](https://vectoralix.com/docs/observability/metrics) How raw logs become rolled-up numbers, what each grain counts, and how fresh the dashboard is at any given window. [Dashboards](https://vectoralix.com/docs/observability/dashboards) A walkthrough of the organization-scoped dashboard widgets and how to read them. ## What this group does NOT cover - Rate-limit enforcement — that is documented in the MCP Deployment group under Rate Limits. Observability records the 429s; the deployment side decides them. - Plan-tier quotas — request-allowance numbers per plan live on the Pricing page, not here. - Internal admin metrics — the platform-admin dashboard is not part of public docs. - A programmatic /api/v1/metrics endpoint — rollups are dashboard-only today; an API surface is on the roadmap. ## What is live today Per-request logging is wired into the MCP protocol endpoint and writes a row to request\_logs on every call. The end-user-facing rollup widgets and the per-row log viewer are still being built — the schema and pipeline shape described in this group are stable, but the in-app surfaces some pages reference are partial. Each page calls out its own status. **Read in order:** Request Logs first (it is the primitive), then Usage Metrics (how those rows become numbers), then Dashboards (what you actually see).