Data access for press and researchers

Everything on the outage tracker is public, and most of it has a machine-readable form alongside the page a person reads. No part of what is described on this page requires an account, an API key, or a signed agreement.

• One service's current reading. Every tracked service has a page at /outages/<slug> and the same answer as JSON at /api/v1/outages/<slug>: the verdict, which regions were read and when, the last 90 days of probe-detected incidents, and 12 months of monthly uptime. Add ?region= to read one region only.
• A combined “stack” reading. A named set of services built on /outages/stack has one page and one JSON answer at /api/v1/outages/stacks/<key>, each member carrying the same reading as its own page plus a roll-up.
• Shared-infrastructure incidents. /outages/internet and its JSON mirror at /api/v1/outages/internet answer whether a named provider (Cloudflare, an AWS region, a CDN) is behind several tracked services going down together, from our own probe correlation, with the measured dependency map and 30 days of declared incidents.
• Cross-catalog latency. /outages/internet-weather and /api/v1/outages/internet-weather publish per-region response-time percentiles across the whole catalog against a 7-day baseline, refreshed hourly.
• Feeds. Every service publishes its own measured-event feed at /outages/<slug>/feed.atom and /feed.xml, and there is one feed across the whole tracker at /outages/feed.atom.
• History. /outages/<slug>/history carries 90 days of per-region availability, a monthly summary table, and the incident log behind the current reading.
• Corrections. /outages/corrections lists every reading we have retracted, what we originally claimed, why it was wrong, and when, the moment a retraction happens.
• Embeds. The embeddable status badge and the internet-weather and research chart images are SVGs anyone can drop into a page, each one carrying its own “Source: RealUptime” line inside the image.

Full request and response shapes for every JSON endpoint above are in the OpenAPI document, under the tag Outages.

None of the JSON endpoints above need an API key, a signup, or a signed agreement: send a plain GET request and read the response. Every one of them also answers with Access-Control-Allow-Origin: *, so a browser-side script on another site can read them directly.

What keeps this open is a per-IP limit rather than a key: 120 requests per minute against any of the JSON endpoints above. Go over it and a request gets a 429 with a Retry-After header naming when to try again, rather than being served stale or silently dropped. Each per-service and per-stack answer is also cache-control'd for two minutes, so polling faster than that returns the same cached reading rather than driving a fresh database read.

There is deliberately no bulk export: no endpoint returns the whole catalog or its full history in one call, and there is no ?all= parameter on the per-service route. Reading the dataset means naming the service (or stack) you want, one request at a time, which is a real limit on how this data can be used, not an oversight. If a research project needs a size or shape of access this does not cover, email us (below) and say what you are trying to do.

Every JSON answer already carries the citation line we ask you to use, in its own source object: a canonical url for the specific reading, and an attribution string reading exactly Measured by RealUptime. Source: <url>. The badge and chart images carry the same idea baked into the SVG itself, as a visible “Source: RealUptime” line. Citing a reading is a matter of keeping that attribution attached to the number, not a request to strip it out.

News style, one line: “According to RealUptime, which independently probes the service from multiple global regions, Example Service was down in the US East and Europe regions for 22 minutes starting at 14:03 UTC (realuptime.io/outages/example-service).”

Academic style, a reference-list entry: RealUptime. (2026). Example Service outage reading [Data set]. Retrieved from https://realuptime.io/outages/example-service

This page states the citation practice already shipped in the product; it is not a signed license. A reading is our own measurement, described in full on the methodology page, and every reading names what it can and cannot prove there, so a citation should carry the same honesty: a reachability check answering “down” is not a claim that every feature of a service failed.

For anything not covered above, larger requests, a question about a specific reading, or anything else, email support@realuptime.io and say what you are working on.