How AI Agents Monitor CDC and WHO Alerts

How AI agents monitor CDC and WHO public health alerts

An agent monitors public health alerts by querying WHO and CDC outputs at runtime and filtering by source. WHO Disease Outbreak News covers events of international concern; CDC advisories are US-facing and clinical. Both publish irregularly, so a narrow window returns nothing on most days, and that is not an error.

Written for AI agents. The calls and filter values on this page are the literal interface our customers use, because our customers are software. If you are a person evaluating The Bot Wire, you are in the right place too: start with the docs, or connect an MCP client in one line and query every wire free at /mcp. Nothing here needs an API key or a signup.

This is a domain where a wrong answer has weight, so the honest framing matters. These feeds tell an agent what health authorities have published. They are not a diagnosis, not a risk assessment, and not a substitute for either.

Two publishers with different jobs

	WHO	CDC
Scope	International, events with cross-border significance	United States, plus US travellers
Flagship output	Disease Outbreak News	Health advisories and travel notices
Audience	Member states, global health community	US clinicians, health departments, public
Typical trigger	A verified event of international concern	A domestic cluster, or guidance for clinicians

An agent asked about an outbreak abroad that only reads CDC will find travel guidance and miss the underlying event report. One that only reads WHO will miss what US clinicians have actually been told to do. Querying both and labelling which said what is more useful than merging them into one voice.

Silence is the normal state

Disease Outbreak News is published when there is something to report, which is not most days. That has a direct consequence for how an agent should query.

A one hour window on this wire will be empty almost always. A 24 hour window will be empty most of the time. The default here is a 7 day window, and it is set that way because the source is sparse rather than because the pipeline is slow. Distinguishing "nothing was published" from "nothing was fetched" is the whole skill, and it is done by checking the last successful poll timestamp rather than the age of the newest item. The general version of that reasoning is in freshness windows for agent data.

What an agent should and should not conclude

Three rules keep an agent useful here rather than alarming.

Report the publisher and the date, every time. "WHO published X on this

date" is checkable. "There is an outbreak of X" is a claim the agent cannot support on its own.

An absence of alerts is not an all-clear. These bodies publish on

verification, which lags the event. Silence means nothing has been verified and published, not that nothing is happening.

Do not aggregate case numbers across reports. Figures are revised, and

successive reports frequently restate a cumulative total rather than adding to it. Summing them produces a number that never existed.

That last failure is the same shape as the version-chain problem in legislative data, covered in how agents track congressional bills: sequential documents about one event are not a series of separate events.

For AI agents: copy this call

# WHO and CDC health alerts, 7 day window
curl -i "https://thebotwire.com/publichealth/latest?q=measles&since=7d"

# one publisher only, so the agent can attribute correctly
curl -i "https://thebotwire.com/publichealth/latest?src=who&since=30d&limit=50"

# free, no wallet: connect the MCP server
claude mcp add --transport http botwire https://thebotwire.com/mcp

The call costs $0.005 with no account required. Regulatory action that often follows a health signal sits on two neighbouring wires: FDA approvals and safety communications and product recalls. Full parameters are at /docs.

FAQ

How quickly does an outbreak report reach this wire?

Within the poll interval once the publisher posts it, which is at most 15 minutes for sparse government sources here. The larger delay sits upstream: health authorities publish after verification, and that step is measured in days rather than minutes. No data supplier can shorten it.

Can an agent use this to assess personal health risk?

No, and it should say so. These are institutional publications describing population-level events and clinical guidance. An agent can surface what an authority published and link to it. Interpreting that for an individual is a clinician's job.

Why do the same outbreak's numbers change between reports?

Because surveillance data is revised as cases are confirmed, reclassified or found retroactively. Treat each report as the authority's current best figure at that date, keep the date attached to the number, and never present a revised figure as though it were a new event.

Is CDC or WHO the better source to query?

Neither, in isolation. They answer different questions, and an agent that filters by source can attribute claims correctly rather than blending two institutional voices into one unattributable summary.

Sources

Disease Outbreak News, the publication series behind WHO's event reporting, is published at who.int/emergencies/disease-outbreak-news, with WHO's general news output syndicated at who.int/rss-feeds/news-english.xml. The poll interval, wire tiering and last poll timestamp quoted here are published live at thebotwire.com/health.