93 sources, each read directly from its publisher
DisruptAPI collects every signal itself, from the authority that issued it: aviation regulators, national weather services, hazard agencies, foreign ministries, health agencies and the status pages of cloud providers. Every signal keeps a link to its source and credits the issuer, and the full register with licences is public at /v1/sources.
Aviation
Weather and hazards
Security, health and strikes
Cloud and internet
History
How a signal is made
- Collect. Each source is polled on its own schedule, from every two minutes for FAA airport status and the main cloud providers to hourly for slow-moving bulletins.
- Normalise. Every item becomes a signal with a type, a subtype, a severity on one five-step scale, start and end times, countries and airports.
- Place. Weather warnings are matched to administrative regions using open boundary data, and every airport inside the region is attached. Where an area name cannot be matched, the warning is kept at country level and says so.
- Track. When a source updates or withdraws an item, the signal is escalated, updated or resolved, and webhooks fire.
How forecasts are made
For each outcome, such as cancellations, check-in failures or websites going down, DisruptAPI looks for past events at the most specific level with at least five measured cases: the same provider and failure layer, then the same failure layer, then all IT outages. It reports a Beta-binomial probability with an 80% interval, the number of cases and example events. Fewer than five cases is flagged as low confidence.
Every outcome is also measured in matched quiet weeks: four windows of the same length with no recorded event, per past event. That base rate is reported next to the probability, with the lift between them, and an outcome whose interval does not clear the base rate is marked as no clear lift and listed last. So far, most kinds of cloud outage do not raise US flight cancellations above a normal week; identity and sign-in outages are the clear exception. We publish that rather than hide it.
Measured outcomes for US flights come from the Bureau of Transportation Statistics: cancellations the airline itself caused, compared with the same weekdays in the four preceding weeks. Weather and other non-IT signals use airport base rates where they exist; analogs for them grow as the warning archive is matched to outcomes.
Rules we keep
- We respect robots.txt and identify ourselves in every request.
- We never get around a login, a block or a CAPTCHA.
- We store facts (times, places, severity, affected components), not copies of articles.
- We skip sources whose licence forbids commercial use.
- We keep each source's required attribution and disclaimers, such as GDACS's "indicative only".
Known limits
- Region-level weather placement covers 38 European countries. A few national services name areas in ways we cannot yet match, and those warnings stay at country level.
- Measured flight outcomes are US-only today. Forecasts outside the US rely on documented past events, and say so.
- DisruptAPI does not report the status of individual flights.
Hear about the disruption before your customers do
Create a key, add the flights, trips and places you want watched, and get a signed webhook whenever a disruption touches one. Try it on a sandbox key first, no card needed.