Ask What Is Happening Right Now: DisasterAWARE's Live Hazards MCP Server

October 23, 2026
Ask What Is Happening Right Now: DisasterAWARE's Live Hazards MCP Server

DisasterAWARE's Live Hazards MCP server connects the AI assistant you already use to DisasterAWARE's live hazard feed: every active hazard worldwide, with its official warning text, its footprint, and the people, hospitals, schools and capital exposed inside it. You ask in plain language: "Which of our twelve ports are inside an active hazard right now, worst first?" The assistant screens your list against every live footprint, ranks what it finds and tells you exactly how fresh the picture is.

It is the live companion of the Historical Hazards MCP server we introduced earlier this month. Connect both and one conversation can answer "what is happening now" and "what happened here before".

What an MCP server is

MCP, the Model Context Protocol, is an open standard for connecting AI assistants to outside data and tools. An MCP server tells the assistant what it can look up and how; the assistant decides which lookups a question needs, runs them and reads the results before it answers.

The Live Hazards MCP server gives an assistant six purpose-built tools over the live feed: one that describes what the feed holds right now and how fresh it is, one that finds and ranks active hazards by type, severity, country, state or province, map window, distance from a point, time or people exposed, one that screens your own sites, routes or areas against every active footprint, one that opens a single hazard with its official text and exposure, one that reports what is new and what changed since a time, and one for counts and exposure totals by type, severity, country, state, urgency, certainty or issuer. You never call these yourself. Knowing they exist simply helps you ask better questions.

What the assistant can reach

The live feed holds every active hazard DisasterAWARE tracks, worldwide: about 1,000 at any moment across some 20 hazard types. Floods, wildfires, tropical cyclones, severe weather, high wind, winter storms, extreme temperatures, earthquakes, volcanoes, landslides, drought, marine and surf hazards, disease outbreaks and smoke, and security incidents.

For each hazard the assistant can read:

  • The official alert. The headline, description and instruction as the issuing agency wrote them, who issued it, how urgent and how certain it is, when it started, when it was last updated, and when it is expected to end or expire.
  • The footprint. The official alert polygon for almost every record, which is what your sites, routes and areas are screened against.
  • The exposure inside it. Modelled population by age group, households, hospitals, schools and capital value, by country and administrative area, and the states, provinces or regions the footprint touches.
  • The clock. The feed is regenerated every few minutes. Every answer states the time of the feed it came from and says so plainly if that feed is older than it should be.

What it does not hold: hazards that have ended, which leave the feed and live on in the historical archive, and the measurements gathered after an event, such as gauge crests, damage surveys and burned acres, which arrive with the historical enhancement.

What makes it different

  • It answers "right now", with a timestamp. Every answer carries the time of the feed generation it came from. If the feed is stale, the assistant says so instead of pretending.
  • It screens your assets, not just the map. Paste a list of sites with coordinates, or a route or area as GeoJSON, add a buffer if you want one, and the assistant returns the hazards intersecting each location, worst first, or the distance to the nearest active footprint.
  • It quotes the official instruction. Ask what the warning covering a city says you should do and you get the issuing agency's own instruction text, not a paraphrase.
  • It counts people and assets, by country and state. Rankings by people, hospitals, schools or capital exposed, grouped the way you ask.
  • It is honest about what a record is. Records are alerts, so one storm can produce several of them, and the assistant says so when it counts.
  • It pairs with the archive. With both servers connected, the assistant can take a hazard that is active now and ask the historical archive what that place has seen before.
  • Setup is documented for the assistants you already use, and sign-in is through your DisasterAWARE account in the browser. You never paste a password or key into the assistant.

Example questions and answers

Every answer below was produced from one generation of the live feed, timestamped 20:32 UTC on 23 October 2026, and checked independently against the DisasterAWARE Active Hazards API. The feed changes every few minutes, so the same questions give different answers today. That is the point.

1. The world right now

Q: What is happening in the world right now, by hazard type and severity?

A: The feed holds 1,077 active hazard records of 20 types touching 177 countries. Wildfire records lead with 439, then floods 195, high wind 113, severe weather 102, severe thunderstorms 44, landslides 39 and extreme temperatures 28, down to five tropical cyclones and four earthquakes. By severity: 224 Warnings, 176 Watches, 572 Advisories and 105 Information records. Almost every record carries an official footprint, 973 were updated in the last 24 hours and more than 300 in the last hour.

Why it matters: a duty officer or analyst gets the shape of the whole day in one answer, with the feed time attached, before drilling into any of it.

2. Where the Warnings are

Q: Which countries have the most active Warnings?

A: The United States with 88 active Warning records, then Ecuador and Kazakhstan with 15 each, Russia with 10, and Peru, the United Kingdom and Canada with 9 each. These are alert records, often one per forecast office or region, so they describe how heavily a country is being warned rather than how many events it has.

Why it matters: corporate security and travel-risk teams see at a glance where official agencies are issuing the most warnings right now.

3. Twelve ports, worst first

Q: Which of these ports are inside an active hazard right now, worst first? (Shanghai, Singapore, Rotterdam, Los Angeles, Houston, Manila, Mumbai, Santos, Durban, Tokyo, Kaohsiung and Miami, with coordinates.)

A: Nine of the twelve sit inside at least one active footprint. Shanghai, Tokyo and Kaohsiung are all inside the alert area of Typhoon Champi. Santos is inside five overlapping severe thunderstorm records from Brazil's national weather service, two of them Warnings: one in force now, the others issued for the next two days. Manila lies inside the warning footprint for Taal volcano, from a Tokyo Volcanic Ash Advisory Centre report of 21 October. Los Angeles is inside a coastal flood advisory for the central and northern California coast that takes effect on 24 October and runs through 29 October. Rotterdam is under a Dutch coastal wind advisory. Houston, Miami and Rotterdam also sit inside continent-scale drought monitor records, which the assistant lists but ranks below the weather warnings. Singapore, Mumbai and Durban are clear. With a 50 km buffer, Singapore picks up a severe weather warning for Peninsular Malaysia whose edge is 29 km away, and Miami picks up four coastal records, among them a coastal flood warning for the Upper Keys and a gale watch for the Florida east coast. Mumbai stays clear: the nearest active footprint, a flood watch in Gujarat, is more than 500 km away.

Why it matters: a logistics control tower or supply-chain team screens its own network against the live picture in seconds, and the "five records at Santos" result shows why the assistant reports alerts rather than counting storms.

4. What the official warning says

Q: What does the official warning covering Santos say I should do?

A: Brazil's National Institute of Meteorology (INMET) issued a Severe Thunderstorm Warning for southern and south-eastern Brazil at 14:58 UTC on 23 October. Urgency Future, certainty Likely, in force from 03:00 UTC on 24 October (just after midnight local time) until 02:59 UTC on 25 October. The record describes very heavy rain of 30 to 60 mm per hour, winds of 60 to 100 km/h and hail, with power outages, crop and structural damage, downed trees and localised flooding expected. Its instruction tells people not to shelter under trees during gusts, to avoid parking near transmission towers or large billboards, to switch off appliances and the main power panel if possible, and gives the civil defence and fire brigade numbers. About 48.9 million people live inside the warning area.

Why it matters: operations and HR teams get the agency's own guidance, with its validity window, for the place they are responsible for, without hunting for the bulletin.

5. Every active tropical cyclone

Q: List every active tropical cyclone and the countries it exposes.

A: Five. Typhoon Champi, north-east of Taiwan, the equivalent of a Category 3 hurricane in the Joint Typhoon Warning Center's forecast issued at 15:00 UTC, with sustained winds of 185 km/h (115 mph) and gusts to 232 km/h (144 mph): about 10.2 million people exposed, almost all in Taiwan and about 55,000 in Japan, with 168 hospitals and 287 schools inside the alert area. Hurricane Vance in the eastern Pacific off Mexico, with a few hundred people in its area. Tropical Storm In-Fa under a Warning near the Marshall Islands, with a few hundred people exposed. Tropical Cyclone Agueda over open ocean in the south-west Indian Ocean, exposing no one. And Post-Tropical Cyclone Josephine off Canada, whose record was minutes from expiring.

Why it matters: insurers, shippers and travel-risk teams get the live storm list with exposure attached, down to the hospitals and schools inside a typhoon's alert area.

Put the live hazard picture in your AI assistant

The Live Hazards MCP server answers what is happening right now, anywhere, from DisasterAWARE's live feed, and screens your own sites and routes against every active footprint.

START FREE TRIAL →TALK TO AN EXPERT →

6. What changed since this morning

Q: What changed in the last six hours?

A: 230 new hazard records and 400 updated ones. The new records were led by wildfire advisories (96), landslide advisories (24), wildfire warnings (23), high wind advisories (15) and severe weather advisories and watches (23 together), plus six new flood warnings. The biggest by people exposed were a new wind alert covering multiple districts of Germany (about 66 million people), a new heavy rain advisory for Tamil Nadu and Karaikal in India (about 41 million) and an updated wind advisory across French departments (about 39 million). The assistant also says what it cannot see: hazards that ended in those six hours have already left the feed, so it does not list them yet.

Why it matters: a duty officer coming on shift gets the delta, not the whole board, and knows which part of the delta the feed cannot show.

7. People under flood warnings, by country

Q: How many people are under flood warnings right now, by country?

A: Colombia leads with about 18.8 million people inside one flood warning covering the north-west of the country, an area holding 1,044 hospitals. A single Central American flood warning exposes about 14.6 million people, almost all of them in Guatemala, with small areas of Honduras and Mexico inside its footprint. The Dominican Republic has about 10.7 million people under a flood warning and Bolivia about 7.8 million. The United States has eight flood warning records from National Weather Service offices, from the Yukon Delta coast to the Florida Keys, exposing about 3 million people between them, most of them under a coastal flood warning for the Southern California coast.

Why it matters: humanitarian, insurance and public-sector teams see where flood warnings touch the most people right now, and can open any record for the agency text and the exposure by administrative area.

8. Live and history in one conversation

Q: Any earthquake or volcano activity in the last 24 hours, and has that area seen this before? (both servers connected)

A: From the live feed: a Warning for Krakatau in Indonesia, an Advisory for Manam in Papua New Guinea, a magnitude 5.7 earthquake under a Watch 61 km west-north-west of Leni in Italy's Aeolian Islands, a magnitude 5.2 nearby off Malfa with about 14,400 people in its area, and a magnitude 5.1 off Noda in Japan with about 39,900 people and two hospitals. From the historical archive: within 100 km of today's Aeolian epicentre there have been four earthquakes of magnitude 5 or greater since 2000, the strongest a magnitude 6.0 in September 2002 north of Santa Flavia and a 5.8 off Ustica in October 2016. Today's 5.7 is the third strongest in that stretch of the Tyrrhenian in a quarter century.

Why it matters: the live server gives the event, the historical server gives the context, and the assistant combines them without you switching tools.

How it is best used

  • Duty officers and security operations centres: the picture at shift start, every new Warning in a region since a time, and the official instruction for the places you are responsible for.
  • Supply chain, logistics and control towers: screen supplier sites, ports, routes and lanes against every active footprint, with a buffer, worst first.
  • Insurance and claims: which insured locations are inside an active hazard now, how many people and assets are exposed, and, with the historical server, what that place has seen before.
  • Travel risk, HR and facilities: what is active around an office, a plant or a traveller, who issued it, and when it expires.
  • Analysts and journalists: rankings and counts by country, state, type and severity, with the feed time attached.

A few habits get the best answers:

  • Start with coverage. "What is in the live feed right now?" tells the assistant what it is working with and how fresh it is.
  • Give coordinates for your sites. A list of names with latitude and longitude, or a GeoJSON line or polygon, is screened exactly. A place name alone is matched against the feed's country and region names.
  • Remember that records are alerts. One real storm can produce several records, one per office or region. Ask for "the biggest" or "the worst" rather than "how many storms".
  • Exposure is modelled. People, hospital and capital figures are estimates for the footprint, not observed impacts.
  • Ask about the ones that are missing. If a hazard you expect is not there, it may have ended; ask the Historical Hazards server.

The MCP server is built for answering questions. To bring the live records themselves into your own systems, use the Active Hazards API, which serves the same feed, or have updates pushed to you by webhook.

How to get started

  1. Add the MCP access permission to your DisasterAWARE account. It is separate from DisasterAWARE web or API access, and the same permission covers the Live and the Historical server. Ask your DisasterAWARE administrator or your PDC account representative to add it.
  2. Add the server to your assistant. Follow the setup steps in the Live Hazards MCP documentation, then sign in with your DisasterAWARE account when the sign-in page opens.
  3. Ask a question. Start with "What is in the live feed right now?" and go from there.

The documentation has setup steps for Claude, Claude Code, ChatGPT and Cursor, the full list of example questions, and tips for accurate answers.

Explore the historical record

Decades of reconstructed, verified catastrophes — browse the Historical Event Intelligence overview, or request sample data for your own portfolio.

HISTORICAL EVENT INTELLIGENCE →REQUEST DATA →

Frequently Asked Questions

What is the DisasterAWARE Live Hazards MCP server?
It is a Model Context Protocol server that connects AI assistants to DisasterAWARE's live hazard feed: about 1,000 active hazards worldwide across some 20 hazard types, each with its official warning text, footprint and modelled exposure. You ask in plain language and the assistant finds, ranks, screens and summarises the live picture, and tells you how fresh it is.

How current is the data?
The feed is regenerated every few minutes, and every answer states the time of the feed generation it came from. If the feed is older than it should be, the assistant says so.

Can I screen my own sites or routes?
Yes. Paste a list of sites with coordinates, or a route or area as GeoJSON, and optionally a buffer distance. The assistant returns the hazards whose footprint intersects each location, worst first, or the distance to the nearest active footprint when none does.

How is it different from the Historical Hazards MCP server?
The live server holds what is active right now and nothing that has ended. The historical server holds the archive of past hazards with the measurements gathered after each event. The same access permission covers both, and with both connected the assistant can answer live and historical questions in one conversation.

Which AI assistants can I use with it?
The documentation has setup steps for Claude, Claude Code, ChatGPT and Cursor. Other assistants that support remote MCP servers with OAuth sign-in can also connect.

Are the exposure numbers observed impacts?
No. Population, household, hospital, school and capital figures are modelled estimates for what lies inside each hazard's footprint. They tell you what is exposed, not what has been damaged.

Related Reading

Sources: DisasterAWARE Active Hazards API, feed generation of 20:32 UTC, 23 October 2026; DisasterAWARE Historical Hazards API, queried 23 October 2026; DisasterAWARE Live Hazards MCP documentation. Official alert text and exposure as carried on each record: Brazil National Institute of Meteorology (INMET), Joint Typhoon Warning Center, National Hurricane Center, Tokyo and Darwin Volcanic Ash Advisory Centres, U.S. National Weather Service, Royal Netherlands Meteorological Institute, U.S. Drought Monitor, USGS.

Get the Weekly Hazard Brief
Curated global hazard intelligence — earthquakes, severe weather, conflict, and cascading risk — delivered to your inbox each week. Free.

More Reading Material