Health status

Live uptime, latency, and incidents for the agntdata platform.

agntdata runs a continuous suite of health probes against every customer-facing resource — the data APIs, the dashboard, the docs, our control-plane API, Supabase, and the upstream dependencies we lean on (Anthropic, Stripe). The current status, recent incidents, and historical uptime are published at status.agntdata.dev.

What’s monitored

Category	Examples	Probe interval
Data APIs	LinkedIn, X, YouTube, TikTok, Facebook, Instagram, Reddit, the agntdata orchestrated APIs	every 30 minutes
Platform	Control-plane API, dashboard app, landing page, docs, Supabase control plane	every 1 minute
Dependency	Anthropic API, Stripe API	every 1 minute

Data API probes use fixture-backed test calls so they exercise the same code path your traffic does without burning extra credits. Platform and dependency probes are lightweight HTTP or Supabase round-trips.

The public JSON surface

If you want to ingest agntdata’s health into your own monitoring (PagerDuty, internal status board, etc.), three unauthenticated endpoints publish the same data the status page renders.

# Current status — one row per active resource
curl https://api.agntdata.dev/public/status/summary

# Recent incidents, newest first
curl "https://api.agntdata.dev/public/status/incidents?limit=50"

# Per-minute history — windowed at 24h, 7d, or 30d
curl "https://api.agntdata.dev/public/status/history?slug=anthropic-api&window=24h"

The summary payload includes the current status (up / degraded / down), last latency, and rolling 24-hour / 7-day / 30-day uptime percentages. The history payload returns per-minute buckets with average latency and error counts so you can chart it yourself.

Incidents

Probe results roll up into api_health_incidents rows whenever a resource goes into a degraded or down state for long enough to count. Each incident has:

• A severity (degraded or down).
• A started_at (when the rollup opened) and ended_at (when probes returned to up; null while ongoing).
• A short summary added by the on-call team when one is published.

The incidents endpoint is the same source the status page renders, so what you see there is what your monitoring will see.

Retention and source

agntdata distinguishes between scheduled sweeps (the recurring probes that power uptime math) and admin sweeps (one-off diagnostic checks the on-call team runs from the admin console). Only scheduled-sweep data feeds the public uptime numbers; admin sweeps are excluded so an ad-hoc diagnostic doesn’t drag down the customer-facing summary.

History is retained for 35 days for scheduled sweeps and 7 days for admin/manual sweeps.

Next steps

Errors

What to do when a request fails — error codes and retry guidance.

Rate limits

Per-plan RPM and how to back off cleanly.