io.github.cyanheads/cdc-health-mcp-server

Official5 toolsLive

by cyanheads · TypeScript

Search and query CDC public health data — mortality, vaccinations, surveillance, behavioral risk.

The MCP server provides tools to search and query CDC public health data—covering mortality, vaccinations, surveillance, and behavioral risk—by accessing the Socrata SODA API. It supports communication via STDIO or Streamable HTTP, exposing 4 tools.

🛠️ Key Features

CDC public health search and querying
Mortality, vaccinations, surveillance, and behavioral risk data coverage
Socrata SODA API integration
4 tools
STDIO or Streamable HTTP transport

🚀 Use Cases

Retrieve mortality metrics from CDC datasets
Look up vaccination information
Query surveillance and behavioral risk data

⚡ Developer Benefits

Works with Model Context Protocol (MCP)
Accesses data through Socrata SODA API
Tool-based interface for CDC data queries
Topics include cdc, health-data, mcp, mcp-server, and typescript

⚠️ Limitations

Limited to CDC data domains listed: mortality, vaccinations, surveillance, and behavioral risk
Described as using Socrata SODA API for data access

Topics

cdchealth-datamcpmcp-servermodel-context-protocolpublic-healthsocratatypescriptcyanheads

Captured live from the server via tools/list.

cdc_discover_datasets

Search the CDC dataset catalog by keyword, category, or tag. Returns IDs, names, truncated descriptions, asset types, column counts, and update timestamps. The catalog also holds charts, maps, stories, files, and links; an entry whose columnCount is 0 is one of those and yields no data from the other tools. Use cdc_get_dataset_schema for the full column list of a chosen dataset.

Parameters7

domain	string	optional	CDC Socrata host to search. "data.cdc.gov" (default) and "chronicdata.cdc.gov" front the same catalog and return the same entries, so switching hosts neither widens nor narrows a search — chronic-disease and small-area collections such as PLACES, the Heart Disease & Stroke Atlas, and Environmental Public Health Tracking are found from either.
query	string	optional	Full-text search across dataset names and descriptions (e.g., "diabetes mortality", "lead exposure children").
category	string	optional	Filter by domain category (e.g., "NNDSS", "Vaccinations", "Behavioral Risk Factors"). Values come from the catalog's own vocabulary and are matched exactly, including case — "Vaccinations" matches 89 entries while "Vaccination" matches none. Call cdc_list_catalog_vocabulary for every value with its entry count rather than guessing at one.
tags	array	optional	Filter by domain tags (e.g., ["covid19", "surveillance"]). Tags widen the search instead of narrowing it — a dataset matches when it carries any one of them, so every tag added returns more results, and an unrecognized tag matches nothing and leaves the result set unchanged. Values match the catalog's own tag vocabulary, case-insensitively; call cdc_list_catalog_vocabulary for the values in use with their entry counts, or read the tags field on any result. To narrow, combine tags with query or category, which intersect with the tag set.
limit	integer	optional	Results to return (default 10, max 100). offset plus limit must not exceed 10000.
offset	integer	optional	Pagination offset for browsing beyond first page (max 9999). offset plus limit must not exceed 10000; both CDC portals hold well under two thousand entries, so offsets near that ceiling page past the end of the catalog.
order	string	optional	Result ordering. "dataset_id" (default) sorts deterministically by each dataset's unique catalog ID — required for stable offset pagination, since consecutive pages form a gap-free, duplicate-free traversal. "relevance" returns best-match ranking for keyword search but is not stably paginable across pages, so walking offsets can skip or repeat datasets.

Raw schema

{
  "type": "object",
  "properties": {
    "domain": {
      "default": "data.cdc.gov",
      "description": "CDC Socrata host to search. \"data.cdc.gov\" (default) and \"chronicdata.cdc.gov\" front the same catalog and return the same entries, so switching hosts neither widens nor narrows a search — chronic-disease and small-area collections such as PLACES, the Heart Disease & Stroke Atlas, and Environmental Public Health Tracking are found from either.",
      "type": "string",
      "enum": [
        "data.cdc.gov",
        "chronicdata.cdc.gov"
      ]
    },
    "query": {
      "description": "Full-text search across dataset names and descriptions (e.g., \"diabetes mortality\", \"lead exposure children\").",
      "type": "string"
    },
    "category": {
      "description": "Filter by domain category (e.g., \"NNDSS\", \"Vaccinations\", \"Behavioral Risk Factors\"). Values come from the catalog's own vocabulary and are matched exactly, including case — \"Vaccinations\" matches 89 entries while \"Vaccination\" matches none. Call cdc_list_catalog_vocabulary for every value with its entry count rather than guessing at one.",
      "type": "string"
    },
    "tags": {
      "description": "Filter by domain tags (e.g., [\"covid19\", \"surveillance\"]). Tags widen the search instead of narrowing it — a dataset matches when it carries any one of them, so every tag added returns more results, and an unrecognized tag matches nothing and leaves the result set unchanged. Values match the catalog's own tag vocabulary, case-insensitively; call cdc_list_catalog_vocabulary for the values in use with their entry counts, or read the tags field on any result. To narrow, combine tags with query or category, which intersect with the tag set.",
      "type": "array",
      "items": {
        "type": "string",
        "description": "Tag value"
      }
    },
    "limit": {
      "default": 10,
      "description": "Results to return (default 10, max 100). offset plus limit must not exceed 10000.",
      "type": "integer",
      "minimum": 1,
      "maximum": 100
    },
    "offset": {
      "default": 0,
      "description": "Pagination offset for browsing beyond first page (max 9999). offset plus limit must not exceed 10000; both CDC portals hold well under two thousand entries, so offsets near that ceiling page past the end of the catalog.",
      "type": "integer",
      "minimum": 0,
      "maximum": 9999
    },
    "order": {
      "default": "dataset_id",
      "description": "Result ordering. \"dataset_id\" (default) sorts deterministically by each dataset's unique catalog ID — required for stable offset pagination, since consecutive pages form a gap-free, duplicate-free traversal. \"relevance\" returns best-match ranking for keyword search but is not stably paginable across pages, so walking offsets can skip or repeat datasets.",
      "type": "string",
      "enum": [
        "dataset_id",
        "relevance"
      ]
    }
  },
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "additionalProperties": false
}

cdc_get_dataset_schema

Fetch the column schema for a CDC dataset — names, data types, descriptions, row count, and last-updated timestamp. Returns the first 100 columns by default; wide datasets continue via column_offset. Get dataset IDs from cdc_discover_datasets.

Parameters4

domain	string	optional	CDC Socrata host to fetch the dataset from. "data.cdc.gov" (default) and "chronicdata.cdc.gov" front the same catalog, so a four-by-four ID resolves on either and the default works whichever host the dataset was found on.
datasetId	string	required	Four-by-four dataset identifier (e.g., "bi63-dtpu"). Obtain from cdc_discover_datasets.
column_limit	integer	optional	Columns to return in this call (default 100, max 500). Every dataset under the default arrives whole; past it the response reports totalCount and a nextOffset to pass back as column_offset. Raise this to pull a wide schema in one call.
column_offset	integer	optional	Index of the first column to return, for continuing past a previous call (default 0). Columns keep the order the dataset declares, so column_offset plus column_limit walks the schema without gaps or repeats. An offset at or past the column count returns an empty window rather than an error.

Raw schema

{
  "type": "object",
  "properties": {
    "domain": {
      "default": "data.cdc.gov",
      "description": "CDC Socrata host to fetch the dataset from. \"data.cdc.gov\" (default) and \"chronicdata.cdc.gov\" front the same catalog, so a four-by-four ID resolves on either and the default works whichever host the dataset was found on.",
      "type": "string",
      "enum": [
        "data.cdc.gov",
        "chronicdata.cdc.gov"
      ]
    },
    "datasetId": {
      "type": "string",
      "pattern": "^[a-z0-9]{4}-[a-z0-9]{4}$",
      "description": "Four-by-four dataset identifier (e.g., \"bi63-dtpu\"). Obtain from cdc_discover_datasets."
    },
    "column_limit": {
      "default": 100,
      "description": "Columns to return in this call (default 100, max 500). Every dataset under the default arrives whole; past it the response reports totalCount and a nextOffset to pass back as column_offset. Raise this to pull a wide schema in one call.",
      "type": "integer",
      "minimum": 1,
      "maximum": 500
    },
    "column_offset": {
      "default": 0,
      "description": "Index of the first column to return, for continuing past a previous call (default 0). Columns keep the order the dataset declares, so column_offset plus column_limit walks the schema without gaps or repeats. An offset at or past the column count returns an empty window rather than an error.",
      "type": "integer",
      "minimum": 0,
      "maximum": 10000
    }
  },
  "required": [
    "datasetId"
  ],
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "additionalProperties": false
}

cdc_list_catalog_vocabulary

List the controlled vocabularies cdc_discover_datasets' category and tags filters are matched against — every domain category and domain tag the CDC catalog publishes, each with the number of entries carrying it. Call it before filtering a search: a value the catalog does not carry matches nothing and returns an empty page, which is indistinguishable from a real value with no results. All 55 categories come back whole; the tag vocabulary runs to roughly 1,600 values, so tags are ranked by entry count and returned one page at a time via tag_limit and tag_offset. Pass filter to narrow both vocabularies to the values whose words contain it.

Parameters4

domain	string	optional	CDC Socrata host to read the vocabulary from. "data.cdc.gov" (default) and "chronicdata.cdc.gov" front the same catalog and publish the same vocabulary, so this selects which host answers, never which values exist.
filter	string	optional	Narrow both vocabularies to the values related to this text. A value matches when every word of the filter appears inside one of its words ("vaccin" reaches "Vaccinations" and "covid-19 vaccination"), or when the whole value appears in the filter. Matching is not fuzzy — a misspelling returns nothing rather than a guess — and a filter under three letters is ignored.
tag_limit	integer	optional	Tags to return in this call (default 50, max 500). Tags are ranked by entry count, so the default page is the most-used end of the vocabulary; the response reports how many matched and a nextOffset while more remain. Categories are never paged — all 55 arrive whole.
tag_offset	integer	optional	Index of the first tag to return, for continuing past a previous call (default 0). Ranking is stable, so tag_offset plus tag_limit walks the vocabulary without gaps or repeats. An offset at or past the number of matching tags returns an empty tag list rather than an error.

Raw schema

{
  "type": "object",
  "properties": {
    "domain": {
      "default": "data.cdc.gov",
      "description": "CDC Socrata host to read the vocabulary from. \"data.cdc.gov\" (default) and \"chronicdata.cdc.gov\" front the same catalog and publish the same vocabulary, so this selects which host answers, never which values exist.",
      "type": "string",
      "enum": [
        "data.cdc.gov",
        "chronicdata.cdc.gov"
      ]
    },
    "filter": {
      "description": "Narrow both vocabularies to the values related to this text. A value matches when every word of the filter appears inside one of its words (\"vaccin\" reaches \"Vaccinations\" and \"covid-19 vaccination\"), or when the whole value appears in the filter. Matching is not fuzzy — a misspelling returns nothing rather than a guess — and a filter under three letters is ignored.",
      "type": "string"
    },
    "tag_limit": {
      "default": 50,
      "description": "Tags to return in this call (default 50, max 500). Tags are ranked by entry count, so the default page is the most-used end of the vocabulary; the response reports how many matched and a nextOffset while more remain. Categories are never paged — all 55 arrive whole.",
      "type": "integer",
      "minimum": 1,
      "maximum": 500
    },
    "tag_offset": {
      "default": 0,
      "description": "Index of the first tag to return, for continuing past a previous call (default 0). Ranking is stable, so tag_offset plus tag_limit walks the vocabulary without gaps or repeats. An offset at or past the number of matching tags returns an empty tag list rather than an error.",
      "type": "integer",
      "minimum": 0,
      "maximum": 10000
    }
  },
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "additionalProperties": false
}

cdc_query_dataset

Execute a SoQL query against any CDC dataset. Supports filtering, aggregation, sorting, full-text search, and field selection. Use cdc_discover_datasets to find dataset IDs and cdc_get_dataset_schema to inspect columns before querying.

Parameters10

domain	string	optional	CDC Socrata host to query. "data.cdc.gov" (default) and "chronicdata.cdc.gov" front the same catalog, so a four-by-four ID returns the same rows from either and the default works whichever host the dataset was found on.
datasetId	string	required	Four-by-four dataset identifier (e.g., "bi63-dtpu"). Obtain from cdc_discover_datasets.
search	string	optional	Full-text search across all text columns. For precise filtering use the where parameter instead.
select	string	optional	SoQL SELECT clause — column names, aliases, or aggregates: "state, sum(deaths) as total_deaths". Omit for all columns. To enumerate distinct values of a column, set select to "{column}, count(*) as count" with group="{column}" and order="count DESC".
where	string	optional	SoQL WHERE clause. Strings must be single-quoted: "state='California' AND year=2020". If a column name matches a SoQL keyword (group, select, where, order, limit, offset, having, search), wrap it in backticks: "`group`='By Year'".
group	string	optional	SoQL GROUP BY clause. Requires aggregate functions in select.
having	string	optional	SoQL HAVING clause. Filters aggregated results.
order	string	optional	SoQL ORDER BY clause. Field name with optional ASC/DESC: "total_deaths DESC". Set one whenever paging with offset: SODA does not order results implicitly, so consecutive offsets without a deterministic order can skip or repeat rows. When the dataset has no natural unique column, Socrata's documented minimum tie-breaker is the system field `:id`, present on every dataset — order=":id".
limit	integer	optional	Max rows to return (default 100, max 5000). Fewer come back when the page would carry the response past its 200,000-character budget, counted over the whole result — the rows as JSON and as the rendered table together; the response says so and gives a nextOffset to resume from.
offset	integer	optional	Row offset for pagination (max 1,000,000). Pair with a deterministic order clause — an offset walk over unordered results can skip or repeat rows.

Raw schema

{
  "type": "object",
  "properties": {
    "domain": {
      "default": "data.cdc.gov",
      "description": "CDC Socrata host to query. \"data.cdc.gov\" (default) and \"chronicdata.cdc.gov\" front the same catalog, so a four-by-four ID returns the same rows from either and the default works whichever host the dataset was found on.",
      "type": "string",
      "enum": [
        "data.cdc.gov",
        "chronicdata.cdc.gov"
      ]
    },
    "datasetId": {
      "type": "string",
      "pattern": "^[a-z0-9]{4}-[a-z0-9]{4}$",
      "description": "Four-by-four dataset identifier (e.g., \"bi63-dtpu\"). Obtain from cdc_discover_datasets."
    },
    "search": {
      "description": "Full-text search across all text columns. For precise filtering use the where parameter instead.",
      "type": "string"
    },
    "select": {
      "description": "SoQL SELECT clause — column names, aliases, or aggregates: \"state, sum(deaths) as total_deaths\". Omit for all columns. To enumerate distinct values of a column, set select to \"{column}, count(*) as count\" with group=\"{column}\" and order=\"count DESC\".",
      "type": "string"
    },
    "where": {
      "description": "SoQL WHERE clause. Strings must be single-quoted: \"state='California' AND year=2020\". If a column name matches a SoQL keyword (group, select, where, order, limit, offset, having, search), wrap it in backticks: \"`group`='By Year'\".",
      "type": "string"
    },
    "group": {
      "description": "SoQL GROUP BY clause. Requires aggregate functions in select.",
      "type": "string"
    },
    "having": {
      "description": "SoQL HAVING clause. Filters aggregated results.",
      "type": "string"
    },
    "order": {
      "description": "SoQL ORDER BY clause. Field name with optional ASC/DESC: \"total_deaths DESC\". Set one whenever paging with offset: SODA does not order results implicitly, so consecutive offsets without a deterministic order can skip or repeat rows. When the dataset has no natural unique column, Socrata's documented minimum tie-breaker is the system field `:id`, present on every dataset — order=\":id\".",
      "type": "string"
    },
    "limit": {
      "default": 100,
      "description": "Max rows to return (default 100, max 5000). Fewer come back when the page would carry the response past its 200,000-character budget, counted over the whole result — the rows as JSON and as the rendered table together; the response says so and gives a nextOffset to resume from.",
      "type": "integer",
      "minimum": 1,
      "maximum": 5000
    },
    "offset": {
      "default": 0,
      "description": "Row offset for pagination (max 1,000,000). Pair with a deterministic order clause — an offset walk over unordered results can skip or repeat rows.",
      "type": "integer",
      "minimum": 0,
      "maximum": 1000000
    }
  },
  "required": [
    "datasetId"
  ],
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "additionalProperties": false
}

cdc_query_wonder

Query CDC WONDER for national US mortality statistics — deaths, population, and crude/age-adjusted death rates — across its five mortality databases, selected with the database input: final underlying-cause data for 1999–2020 (the default) or 2018–2024, provisional data running from 2018 through the current year, and two multiple-cause databases covering the same two eras. Break results out by year, age group, sex, and/or race, and filter by ICD-10 cause of death, sex, age group, or year range; on a multiple-cause database, mcd_icd10 additionally matches a cause listed anywhere on the death certificate rather than only the one certified as underlying. Each database holds a different span of years (1999–2026 across all of them) and a request whose year_range falls outside the selected one's span is rejected with that span named. WONDER is a separate CDC system from the Socrata datasets the other cdc_* tools query. Data is national only — sub-national (state/county) breakdowns are not available through the API (CDC vital-statistics policy). Cause of death is a filter, not a grouping. Some measure cells come back as a CDC status token rather than a number — "Suppressed" (withheld for confidentiality), "Unreliable" (a rate from fewer than 20 deaths), or "Not Applicable" (no population denominator); those cells read null in rows and each one is listed in cellNotes with its token. CDC also drops whole rows before sending the table — strata with zero deaths, and strata whose death count is suppressed — so a stratum can be missing from rows entirely; messages carries CDC's statement whenever that happened. Each response is bounded by a 200,000-character budget counted over the whole result, so a broad grouping — one can run past two thousand rows — comes back a page at a time: the response reports the table's totalCount and a nextOffset to continue from, and limit takes smaller pages. Paging shapes the response only — WONDER is asked once either way, and the figures, caveats and hidden-row notices are the same on every page. CDC rejects requests made less than 15 seconds apart across all five databases, so calls are spaced automatically: calls made while another is running wait their turn and run one after another, each queued call adding about 16 seconds plus its own query time before it returns.

Parameters9

database	string	optional	Which WONDER mortality database to query. "underlying_1999_2020" (D76) is final data for 1999–2020 and the default. "provisional" (D176) runs 2018 through the current year, updated weekly, and returns the most recent years labelled e.g. "2025 (provisional)". "underlying_2018_2024" (D158) is settled — not provisional — data for 2018–2024. "multiple_1999_2020" (D77) and "multiple_2018_2024" (D157) record every cause listed on the death certificate; without an mcd_icd10 filter they return the same figures as the underlying-cause database for the same era, so pick one only to use that filter. The two 1999–2020 databases report race in CDC's four bridged groups; the other three use the six single-race groups — figures broken out by race are not comparable between the two families.
group_by	array	optional	Dimensions to break results out by (1–4, each at most once), in output-column order — e.g. ["year"], ["year","sex"], ["age_group","race"]. Results are always national. Cause of death is a filter (cause_icd10), not a grouping. "race" resolves to whichever race vocabulary the selected database uses — four bridged groups (Asian and Pacific Islander combined) on the 1999–2020 databases, six single-race categories on the others, one of them "More than one race" — so a race series from one family cannot be spliced onto one from the other.
cause_icd10	any	optional	Filter to ICD-10 underlying causes of death — the single condition CDC certified as having started the chain of events leading to death. Takes one code or range, or a list of them for a cause defined as a code set, e.g. drug overdose as ["X40","X41","X42","X43","X44","X60","X61","X62","X63","X64","X85","Y10","Y11","Y12","Y13","Y14"]. Omit for all causes. Accepted by every database. "999--999" is not an ICD-10 code but CDC's own marker for deaths whose cause it is still withholding under the provisional database's six-month reporting lag; it counts that backlog, and only the "provisional" database offers it.
mcd_icd10	any	optional	Filter to deaths with any of these ICD-10 codes recorded anywhere on the death certificate, whether or not it was the underlying cause — e.g. "died with a respiratory condition listed", a population no underlying-cause query can produce. Takes one code or range, or a list of them, e.g. opioid involvement as ["T40.0","T40.1","T40.2","T40.3","T40.4","T40.6"]. Valid only when database is "multiple_1999_2020", "multiple_2018_2024", or "provisional"; the other databases record only the underlying cause and reject it. "999--999", the withheld-cause marker described under cause_icd10, is offered here too but only by "provisional". Combines with cause_icd10, which keeps meaning the underlying cause: a death must match both filters. Omit for all causes.
sex	string	optional	Filter by sex.
age_groups	array	optional	Restrict to deaths in any of the listed age groups — e.g. ["25-34","35-44"] covers both; a repeated group counts once. "1" is the under-1-year group. "NS" is the group CDC puts a death in when the age was not recorded; it is not covered by any of the ten-year groups, so a filter listing all eleven of those still leaves those deaths out and returns fewer deaths than the same query unfiltered. List "NS" alongside them to match an unfiltered total, or on its own to count them. Omit for all ages, which includes them.
year_range	object	optional	Inclusive year range; from must not be later than to. These bounds span every database (1999–2026); the years the selected one actually holds are narrower, and a range outside them is rejected with that database's span named. Omit for all years the database holds.
limit	integer	optional	Rows to return from the table CDC sent (1–5000). Omit to take as many as fit. Either way fewer come back when the page would carry the response past its 200,000-character budget, counted over the whole result — the rows, their cell notes, and the caveats and messages, as JSON and as the rendered table together; the response says so and gives a nextOffset to resume from. WONDER's request carries no limit of its own, so this pages a table already fetched in full rather than narrowing the query: the deaths, rates, caveats and hidden-row notices are the same whichever page is read.
offset	integer	optional	Index of the first row to return, for continuing past a previous call (default 0, max 10,000). Rows keep the order CDC returned them in, which is stable for a given query, so offset plus limit walks the table without gaps or repeats. An offset at or past the row total returns an empty page rather than an error.

Raw schema

{
  "type": "object",
  "properties": {
    "database": {
      "default": "underlying_1999_2020",
      "description": "Which WONDER mortality database to query. \"underlying_1999_2020\" (D76) is final data for 1999–2020 and the default. \"provisional\" (D176) runs 2018 through the current year, updated weekly, and returns the most recent years labelled e.g. \"2025 (provisional)\". \"underlying_2018_2024\" (D158) is settled — not provisional — data for 2018–2024. \"multiple_1999_2020\" (D77) and \"multiple_2018_2024\" (D157) record every cause listed on the death certificate; without an mcd_icd10 filter they return the same figures as the underlying-cause database for the same era, so pick one only to use that filter. The two 1999–2020 databases report race in CDC's four bridged groups; the other three use the six single-race groups — figures broken out by race are not comparable between the two families.",
      "type": "string",
      "enum": [
        "underlying_1999_2020",
        "provisional",
        "underlying_2018_2024",
        "multiple_1999_2020",
        "multiple_2018_2024"
      ]
    },
    "group_by": {
      "default": [
        "year"
      ],
      "description": "Dimensions to break results out by (1–4, each at most once), in output-column order — e.g. [\"year\"], [\"year\",\"sex\"], [\"age_group\",\"race\"]. Results are always national. Cause of death is a filter (cause_icd10), not a grouping. \"race\" resolves to whichever race vocabulary the selected database uses — four bridged groups (Asian and Pacific Islander combined) on the 1999–2020 databases, six single-race categories on the others, one of them \"More than one race\" — so a race series from one family cannot be spliced onto one from the other.",
      "minItems": 1,
      "maxItems": 4,
      "type": "array",
      "items": {
        "type": "string",
        "enum": [
          "year",
          "age_group",
          "sex",
          "race"
        ]
      }
    },
    "cause_icd10": {
      "description": "Filter to ICD-10 underlying causes of death — the single condition CDC certified as having started the chain of events leading to death. Takes one code or range, or a list of them for a cause defined as a code set, e.g. drug overdose as [\"X40\",\"X41\",\"X42\",\"X43\",\"X44\",\"X60\",\"X61\",\"X62\",\"X63\",\"X64\",\"X85\",\"Y10\",\"Y11\",\"Y12\",\"Y13\",\"Y14\"]. Omit for all causes. Accepted by every database. \"999--999\" is not an ICD-10 code but CDC's own marker for deaths whose cause it is still withholding under the provisional database's six-month reporting lag; it counts that backlog, and only the \"provisional\" database offers it.",
      "anyOf": [
        {
          "type": "string",
          "const": ""
        },
        {
          "type": "string",
          "const": "999--999"
        },
        {
          "type": "string",
          "pattern": "^[A-Z][0-9]{2}(\\.[0-9]+)?(-[A-Z][0-9]{2}(\\.[0-9]+)?)?$",
          "description": "ICD-10 underlying-cause code or range. WONDER takes a node of its ICD-10 tree — a chapter such as \"A00-B99\" (infectious) or \"V01-Y89\" (external causes), a block such as \"X40-X49\" (accidental poisoning) or \"C00-C97\" (malignant neoplasms), or a single code such as \"I21\" — and rejects any other span, e.g. \"X40-X44\", naming it in the error. List the codes (or blocks) to cover a span that is not a tree node."
        },
        {
          "minItems": 1,
          "maxItems": 50,
          "type": "array",
          "items": {
            "anyOf": [
              {
                "type": "string",
                "const": "999--999"
              },
              {
                "type": "string",
                "pattern": "^[A-Z][0-9]{2}(\\.[0-9]+)?(-[A-Z][0-9]{2}(\\.[0-9]+)?)?$",
                "description": "ICD-10 underlying-cause code or range. WONDER takes a node of its ICD-10 tree — a chapter such as \"A00-B99\" (infectious) or \"V01-Y89\" (external causes), a block such as \"X40-X49\" (accidental poisoning) or \"C00-C97\" (malignant neoplasms), or a single code such as \"I21\" — and rejects any other span, e.g. \"X40-X44\", naming it in the error. List the codes (or blocks) to cover a span that is not a tree node."
              }
            ],
            "description": "One list entry: an ICD-10 code or range, or the withheld-cause marker."
          },
          "description": "A list of 1–50 entries, each in the single-value form, matched as a union: a death counts once when it matches any of them, so a code set such as X40–X44 plus X60–X64 is one series with one set of rates. Repeated entries are ignored."
        }
      ]
    },
    "mcd_icd10": {
      "description": "Filter to deaths with any of these ICD-10 codes recorded anywhere on the death certificate, whether or not it was the underlying cause — e.g. \"died with a respiratory condition listed\", a population no underlying-cause query can produce. Takes one code or range, or a list of them, e.g. opioid involvement as [\"T40.0\",\"T40.1\",\"T40.2\",\"T40.3\",\"T40.4\",\"T40.6\"]. Valid only when database is \"multiple_1999_2020\", \"multiple_2018_2024\", or \"provisional\"; the other databases record only the underlying cause and reject it. \"999--999\", the withheld-cause marker described under cause_icd10, is offered here too but only by \"provisional\". Combines with cause_icd10, which keeps meaning the underlying cause: a death must match both filters. Omit for all causes.",
      "anyOf": [
        {
          "type": "string",
          "const": ""
        },
        {
          "type": "string",
          "const": "999--999"
        },
        {
          "type": "string",
          "pattern": "^[A-Z][0-9]{2}(\\.[0-9]+)?(-[A-Z][0-9]{2}(\\.[0-9]+)?)?$",
          "description": "ICD-10 code or range, same form as cause_icd10 — a chapter such as \"S00-T98\" (injury and poisoning, a chapter the underlying-cause finder does not list), a block such as \"J09-J18\" (influenza and pneumonia), or a single code such as \"T40.1\" (heroin)."
        },
        {
          "minItems": 1,
          "maxItems": 50,
          "type": "array",
          "items": {
            "anyOf": [
              {
                "type": "string",
                "const": "999--999"
              },
              {
                "type": "string",
                "pattern": "^[A-Z][0-9]{2}(\\.[0-9]+)?(-[A-Z][0-9]{2}(\\.[0-9]+)?)?$",
                "description": "ICD-10 code or range, same form as cause_icd10 — a chapter such as \"S00-T98\" (injury and poisoning, a chapter the underlying-cause finder does not list), a block such as \"J09-J18\" (influenza and pneumonia), or a single code such as \"T40.1\" (heroin)."
              }
            ],
            "description": "One list entry: an ICD-10 code or range, or the withheld-cause marker."
          },
          "description": "A list of 1–50 entries, each in the single-value form, matched as a union: a death counts once when it matches any of them, so a code set such as X40–X44 plus X60–X64 is one series with one set of rates. Repeated entries are ignored."
        }
      ]
    },
    "sex": {
      "default": "all",
      "description": "Filter by sex.",
      "type": "string",
      "enum": [
        "all",
        "male",
        "female"
      ]
    },
    "age_groups": {
      "description": "Restrict to deaths in any of the listed age groups — e.g. [\"25-34\",\"35-44\"] covers both; a repeated group counts once. \"1\" is the under-1-year group. \"NS\" is the group CDC puts a death in when the age was not recorded; it is not covered by any of the ten-year groups, so a filter listing all eleven of those still leaves those deaths out and returns fewer deaths than the same query unfiltered. List \"NS\" alongside them to match an unfiltered total, or on its own to count them. Omit for all ages, which includes them.",
      "type": "array",
      "items": {
        "type": "string",
        "enum": [
          "1",
          "1-4",
          "5-14",
          "15-24",
          "25-34",
          "35-44",
          "45-54",
          "55-64",
          "65-74",
          "75-84",
          "85+",
          "NS"
        ]
      }
    },
    "year_range": {
      "description": "Inclusive year range; from must not be later than to. These bounds span every database (1999–2026); the years the selected one actually holds are narrower, and a range outside them is rejected with that database's span named. Omit for all years the database holds.",
      "type": "object",
      "properties": {
        "from": {
          "type": "integer",
          "minimum": 1999,
          "maximum": 2026,
          "description": "First year (1999–2026 across all databases; the selected one holds a narrower span)."
        },
        "to": {
          "type": "integer",
          "minimum": 1999,
          "maximum": 2026,
          "description": "Last year (1999–2026 across all databases; the selected one holds a narrower span)."
        }
      },
      "required": [
        "from",
        "to"
      ]
    },
    "limit": {
      "description": "Rows to return from the table CDC sent (1–5000). Omit to take as many as fit. Either way fewer come back when the page would carry the response past its 200,000-character budget, counted over the whole result — the rows, their cell notes, and the caveats and messages, as JSON and as the rendered table together; the response says so and gives a nextOffset to resume from. WONDER's request carries no limit of its own, so this pages a table already fetched in full rather than narrowing the query: the deaths, rates, caveats and hidden-row notices are the same whichever page is read.",
      "type": "integer",
      "minimum": 1,
      "maximum": 5000
    },
    "offset": {
      "default": 0,
      "description": "Index of the first row to return, for continuing past a previous call (default 0, max 10,000). Rows keep the order CDC returned them in, which is stable for a given query, so offset plus limit walks the table without gaps or repeats. An offset at or past the row total returns an empty page rather than an error.",
      "type": "integer",
      "minimum": 0,
      "maximum": 10000
    }
  },
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "additionalProperties": false
}

README.md20 KBView on GitHub

@cyanheads/cdc-health-mcp-server

Search and query CDC public health data — mortality, vaccinations, surveillance, behavioral risk (Socrata SODA API) via MCP. STDIO or Streamable HTTP.

5 Tools • 2 Resources • 1 Prompt

MCP SDK TypeScript

Public Hosted Server: https://cdc.caseyjhand.com/mcp

Overview

CDC public health data — the Socrata-based CDC Open Data portal, plus CDC WONDER, a separate CDC system for national mortality statistics. Search the catalog, inspect dataset schemas, and run SoQL queries across vaccination, surveillance, and behavioral-risk data, or query WONDER for deaths, population, and death rates by year, age, sex, and race. Runs as a stdio process, a local Streamable HTTP server, or the public hosted endpoint above.

Tools

Tool	Description
`cdc_discover_datasets`	Search the CDC dataset catalog by keyword, category, or tag
`cdc_get_dataset_schema`	Fetch column schema, row count, and metadata for a dataset
`cdc_list_catalog_vocabulary`	List the catalog's category and tag values with their entry counts
`cdc_query_dataset`	Execute SoQL queries — filter, aggregate, sort, full-text search, select fields
`cdc_query_wonder`	Query CDC WONDER for national mortality, population, and death rates across five databases

Resources

Resource	Description
`cdc://datasets`	Top 50 most-viewed catalog entries, for orientation
`cdc://datasets/{datasetId}`	Dataset metadata and the first 100 columns for a specific dataset

Both resources mirror data also reachable via cdc_discover_datasets and cdc_get_dataset_schema, for clients that surface resources but not tools.

Prompts

Prompt	Description
`analyze_health_trend`	Guided workflow for investigating a public health question across CDC data

Capability reference

`cdc_discover_datasets` _tool

domain selects data.cdc.gov (default) or chronicdata.cdc.gov — both front the same catalog, so switching hosts neither widens nor narrows a search
query, category, and tags filters — tags union (a dataset matches on any one tag, so each tag added widens the result set), while query and category intersect with the tag set
Up to 100 results per page (default 10); offset is capped at 9999, and offset + limit must not exceed 10,000 — Socrata's catalog ceiling
order: dataset_id (default) sorts deterministically for stable pagination; relevance ranks by best match but is not stably paginable across pages
Each result carries columnCount — a value of 0 marks a non-tabular asset (chart, map, story, file, or href) that yields no data from the other tools; assetType is descriptive only
description is converted to plain text before it is cut to 300 characters, so the budget buys visible text rather than the HTML tags catalog entries arrive wrapped in
Enrichment carries totalCount and appliedFilters; a notice distinguishes an offset past the end of the result set from a search that matched nothing, and resolves a category/tags value that matched nothing against the catalog vocabulary — category: "Vaccination" comes back naming "Vaccinations" (89 datasets) rather than advising a broader search. The vocabulary is read only on that branch, so an ordinary search costs no extra request

`cdc_get_dataset_schema` _tool

Accepts a four-by-four datasetId (e.g. bi63-dtpu) and the same domain enum as the other Socrata tools
Returns the first 100 columns by default (column_limit, max 500) — catalog schemas run 3 to 322 columns, so ordinary datasets arrive whole; wider ones report totalCount, truncated, and nextOffset to pass back as column_offset
A column_offset at or past the column count returns an empty window rather than an error
Fails with not_queryable when the ID names a non-tabular catalog asset, rather than returning an empty column list
rowCount prefers a live count(*) fetched alongside the metadata; rowCountSource says live or cached, since Socrata's cached figure is built once and can understate an actively-updated dataset by a third or more. A failed count falls back to the cached figure and never fails the schema response
description is returned in full as plain text — markup stripped, entity references decoded; only cdc_discover_datasets truncates it

`cdc_list_catalog_vocabulary` _tool

The controlled vocabularies cdc_discover_datasets' category and tags filters are matched against — a value the catalog does not carry matches nothing, which is indistinguishable from a real value with no results
All 55 categories return whole (~3 KB); the tag vocabulary runs to 1,583 values, so tags are ranked by entry count and paged with tag_limit (default 50, max 500) and tag_offset
The service asks the tag endpoint for the whole vocabulary explicitly — left to its default it returns 100 values and reports resultSetSize: 100 beside them, so the under-count reads as complete
filter narrows both vocabularies before the page is cut, matching on whole words in either direction ("vaccin" reaches Vaccinations and covid-19 vaccination). Not fuzzy — a misspelling returns nothing rather than a guess
Enrichment carries vocabularySize, the matched categoryCount/tagCount, and truncated/shown/cap/nextOffset; truncationCeiling bounds every omitted tag, since the list is ranked by the same count
Both hosts publish the same vocabulary — measured live, data.cdc.gov and chronicdata.cdc.gov return the identical 55 categories and 1,583 tags

`cdc_query_dataset` _tool

Full SoQL support — select, where, group, having, order, plus full-text search across text columns
Up to 5,000 rows per request (default 100); offset capped at 1,000,000
truncated is measured by an over-fetch probe (one row past the limit), never guessed from the row count; the whole response — structuredContent and content[] together — is bounded by a 200,000-character budget, so a wide page can end short of limit with a nextOffset
An empty page at offset > 0 is diagnosed with one probe at offset 0: the notice says whether the offset ran past the end or the query matches nothing, and names both causes if the probe fails
effectiveQuery echoes the SoQL clauses sent in their original text, not URL-encoded, so a clause can be copied back into the parameter it came from
Fails with not_queryable when every returned row carries no fields and the asset reports no columns — a chart or map ID, which Socrata answers 200 with a body of empty objects. When the asset does have columns, the same shape is a null-only projection and comes back as a success with a notice
All response values are strings (SODA v2.1) — parse per the column's dataType from the schema

`cdc_query_wonder` _tool

database selects which of five mortality databases answers the query:

Value	CDC database	Years	Race groups	`mcd_icd10`
`underlying_1999_2020` (default)	D76 — Underlying Cause of Death	1999–2020	4 bridged	—
`provisional`	D176 — Provisional Mortality Statistics	2018 → current year	6 single-race	yes
`underlying_2018_2024`	D158 — Underlying Cause of Death, Single Race	2018–2024	6 single-race	—
`multiple_1999_2020`	D77 — Multiple Cause of Death	1999–2020	4 bridged	yes
`multiple_2018_2024`	D157 — Multiple Cause of Death, Single Race	2018–2024	6 single-race	yes

group_by: 1–4 of year, age_group, sex, race, each at most once; national totals only — no sub-national breakdown at any setting
cause_icd10 and mcd_icd10 take one ICD-10 code or range, or a list of up to 50 matched as one union — e.g. the drug-overdose set ["X40","X41","X42","X43","X44","X60","X61","X62","X63","X64","X85","Y10","Y11","Y12","Y13","Y14"] returns one series with one set of rates. A range must be a chapter or block of WONDER's ICD-10 tree (X40-X49); any other span (X40-X44) is rejected, so list its codes instead
mcd_icd10 matches a cause recorded anywhere on the death certificate rather than only the underlying cause; accepted only by provisional, multiple_1999_2020, and multiple_2018_2024 — the others reject it
age_groups must include "NS" (age not recorded) to match an unfiltered total; a year_range outside the selected database's span is rejected with that span named
Measure cells CDC withholds or flags (Suppressed, Unreliable, Not Applicable) read null in rows and are named per cell in cellNotes; whole rows CDC hides (zero or suppressed deaths) are absent from rows with no gap marker — check messages
Each response is bounded by a 200,000-character budget over structuredContent and content[] together, so a large grouping comes back a page at a time with an exact totalCount, truncated, and a nextOffset to continue from; limit (max 5,000) takes smaller pages and offset (max 10,000) resumes
Consecutive requests are spaced 16 seconds automatically — CDC rejects anything sent less than 15 seconds after the prior response finished, measured across all five databases. Concurrent calls queue and run one at a time, each queued call adding about 16 seconds
CDC's caveats and notices keep their methodology links as Markdown links

`cdc://datasets` _resource

Top 50 CDC catalog entries by popularity, each carrying assetType and columnCount for orientation
columnCount: 0 marks a non-tabular entry (chart, map, story, file, or href); use cdc_discover_datasets for full catalog search with filtering and pagination

`cdc://datasets/{datasetId}` _resource

Dataset metadata plus the first 100 columns as application/json; datasetId is a four-by-four identifier from cdc_discover_datasets
Carries the dataset's total columnCount and a truncated flag; wider schemas continue via cdc_get_dataset_schema with column_offset
Takes no query-parameter selector — an RFC 6570 {?column_limit,column_offset} template would stop the bare cdc://datasets/{datasetId} form from matching at all

`analyze_health_trend` _prompt

Arguments: topic required; timeRange and geography optional
Returns one user message that routes the question to CDC WONDER (national mortality, 1999–current, ICD-10-filterable) or the Socrata catalog (everything else), then walks discover → inspect → baseline query → compare → synthesize
Routing is prose for the reader to act on — the handler does not classify the topic itself

Features

Built on @cyanheads/mcp-ts-core: stdio and Streamable HTTP transports, pluggable auth (none / jwt / oauth), swappable storage (in-memory, filesystem, Supabase, Cloudflare KV/R2/D1), structured logging with optional OpenTelemetry tracing.

CDC-specific:

Wraps the Socrata SODA API v2.1 (the CDC Open Data portal, ~1,080 datasets) — no auth required, optional app token for higher rate limits
Adds CDC WONDER mortality access (cdc_query_wonder) — a separate XML-over-HTTP CDC system, spanning five mortality databases from 1999 through the current year
Discovery-first workflow for a heterogeneous catalog — discover, inspect schema, then query
Two Socrata hosts via the domain input (data.cdc.gov, chronicdata.cdc.gov), allowlisted at the schema level — both front one tenant, so assets like PLACES and the Heart Disease & Stroke Atlas are reachable from either
Conservative request spacing for both APIs — no rate-limit headers from Socrata, and CDC WONDER requests are spaced 16 seconds apart automatically

Agent-friendly output:

Pagination and truncation disclosed on every tool — totalCount/truncated/nextOffset (or shown/cap) rather than a bare row count, so an agent can tell a complete result from a page of one
Typed error contracts with a recovery hint on every declared reason (e.g. not_queryable, page_out_of_range) — actionable next steps, not just an error code
Upstream data gaps stay visible rather than silently dropped — CDC's status tokens (Suppressed, Unreliable, Not Applicable) are named per cell in cellNotes, and hidden-row notices surface in messages
effectiveQuery echoes the exact query sent (SoQL clauses or a WONDER summary), so a result is reproducible and a clause can be copied back into its parameter

Getting started

Public Hosted Instance

A public instance is available at https://cdc.caseyjhand.com/mcp — no installation required. Point any MCP client at it via Streamable HTTP:

json

{
  "mcpServers": {
    "cdc-health-mcp-server": {
      "type": "streamable-http",
      "url": "https://cdc.caseyjhand.com/mcp"
    }
  }
}

Self-Hosted / Local

Add the following to your MCP client configuration file.

json

{
  "mcpServers": {
    "cdc-health-mcp-server": {
      "type": "stdio",
      "command": "bunx",
      "args": ["@cyanheads/cdc-health-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_LOG_LEVEL": "info"
      }
    }
  }
}

Or with npx (no Bun required):

json

{
  "mcpServers": {
    "cdc-health-mcp-server": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@cyanheads/cdc-health-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_LOG_LEVEL": "info"
      }
    }
  }
}

Or with Docker:

json

{
  "mcpServers": {
    "cdc-health-mcp-server": {
      "type": "stdio",
      "command": "docker",
      "args": ["run", "-i", "--rm", "-e", "MCP_TRANSPORT_TYPE=stdio", "ghcr.io/cyanheads/cdc-health-mcp-server:latest"]
    }
  }
}

For Streamable HTTP, set the transport and start the server:

MCP_TRANSPORT_TYPE=http MCP_HTTP_PORT=3010 bun run start:http
# Server listens at http://localhost:3010/mcp

Prerequisites

Bun v1.4.0 or higher.
Optional: Socrata app token for higher rate limits.

Installation

Clone the repository:

git clone https://github.com/cyanheads/cdc-health-mcp-server.git

Navigate into the directory:

cd cdc-health-mcp-server

Install dependencies:

bun install

Configure environment:

cp .env.example .env
# edit .env and set optional overrides

Configuration

Variable	Description	Default
`MCP_TRANSPORT_TYPE`	Transport: `stdio` or `http`	`stdio`
`MCP_HTTP_PORT`	HTTP server port	`3010`
`MCP_SESSION_MODE`	HTTP session posture: `stateful`, `stateless`, or `auto`. `src/index.ts` declares `stateless`; this variable overrides it	`stateless`
`MCP_AUTH_MODE`	Authentication: `none`, `jwt`, or `oauth`	`none`
`MCP_LOG_LEVEL`	Log level (`debug`, `info`, `warning`, `error`, etc.)	`info`
`LOGS_DIR`	Directory for log files (Node.js only)	`<project-root>/logs`
`STORAGE_PROVIDER_TYPE`	Storage backend: `in-memory`, `filesystem`, `supabase`, `cloudflare-kv/r2/d1`	`in-memory`
`CDC_APP_TOKEN`	Socrata app token for higher rate limits	—
`CDC_BASE_URL`	SODA host for requests that name no `domain` — in practice only the `cdc://datasets/{datasetId}` resource. The three Socrata tools always send a `domain` (`data.cdc.gov` by default), which overrides this	`https://data.cdc.gov`
`CDC_CATALOG_URL`	Base URL for Socrata Discovery API	`https://api.us.socrata.com/api/catalog/v1`
`OTEL_ENABLED`	Enable OpenTelemetry instrumentation (spans, metrics, completion logs)	`false`

See .env.example for the full list of optional overrides.

Running the server

Local development

Build and run the production version:

# One-time build
bun run rebuild

# Run the built server
bun run start:http
# or
bun run start:stdio

Run checks and tests:

bun run devcheck  # Lints, formats, type-checks, and more
bun run test      # Runs the test suite

Docker

docker build -t cdc-health-mcp-server .
docker run --rm -p 3010:3010 cdc-health-mcp-server

The Dockerfile defaults to HTTP transport, stateless session mode, and logs to /var/log/cdc-health-mcp-server. OpenTelemetry peer dependencies are installed by default — build with --build-arg OTEL_ENABLED=false to omit them.

Project structure

Directory	Purpose
`src/index.ts`	`createApp()` entry point — registers tools/resources/prompts and inits services.
`src/config`	Server-specific environment variable parsing and validation with Zod.
`src/mcp-server/tools`	Tool definitions (`*.tool.ts`). Four CDC data tools.
`src/mcp-server/resources`	Resource definitions. Catalog overview and dataset detail.
`src/mcp-server/prompts`	Prompt definitions. Health trend analysis workflow.
`src/services/socrata`	Socrata SODA API service layer — HTTP client, catalog search, metadata, queries.
`src/services/wonder`	CDC WONDER service layer — XML request builder and response parser.
`src/utils`	Shared helpers, including `escapeTableCell` for `format()` output.
`tests/`	Unit and integration tests mirroring `src/`.

Development guide

See CLAUDE.md for development guidelines and architectural rules. The short version:

Handlers throw, framework catches — no try/catch in tool logic
Use ctx.log for logging, ctx.state for storage
Register new tools and resources in the createApp() arrays
Wrap external API calls: validate raw → normalize to domain type → return output schema; never fabricate missing fields

Contributing

Issues are welcome. Run checks and tests before submitting:

bun run devcheck
bun run test

License

Apache-2.0 — see LICENSE for details.

Install

Tools (5)

cdc_discover_datasetscdc_get_dataset_schemacdc_list_catalog_vocabularycdc_query_datasetcdc_query_wonder

Configuration

Environment variables

CDC_APP_TOKEN

Socrata app token for higher rate limits.

MCP_LOG_LEVELdefault info

Sets the minimum log level for output (e.g., 'debug', 'info', 'warn').

MCP_HTTP_HOSTdefault 127.0.0.1

The hostname for the HTTP server.

MCP_HTTP_PORTdefault 3010

The port to run the HTTP server on.

MCP_HTTP_ENDPOINT_PATHdefault /mcp

The endpoint path for the MCP server.

MCP_AUTH_MODEdefault none

Authentication mode to use: 'none', 'jwt', or 'oauth'.

Add to VS Code Add to Cursor

claude_desktop_config.json

json

{
  "mcpServers": {
    "cdc-health-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "@cyanheads/cdc-health-mcp-server@0.9.2",
        "run",
        "start:stdio"
      ],
      "env": {
        "CDC_APP_TOKEN": "<YOUR_CDC_APP_TOKEN>",
        "MCP_LOG_LEVEL": "info"
      }
    }
  }
}

🛠️ Key Features

🚀 Use Cases

⚡ Developer Benefits

⚠️ Limitations

Topics

cdc_discover_datasets

cdc_get_dataset_schema

cdc_list_catalog_vocabulary

cdc_query_dataset

cdc_query_wonder

@cyanheads/cdc-health-mcp-server

Overview

Tools

Resources

Prompts

Capability reference

cdc_discover_datasets tool

cdc_get_dataset_schema tool

cdc_list_catalog_vocabulary tool

cdc_query_dataset tool

cdc_query_wonder tool

cdc://datasets resource

cdc://datasets/{datasetId} resource

analyze_health_trend prompt

Features

Getting started

Public Hosted Instance

Self-Hosted / Local

Prerequisites

Installation

Configuration

Running the server

Local development

Docker

Project structure

Development guide

Contributing

License

Install

Configuration

Related servers

`cdc_discover_datasets` _tool

`cdc_get_dataset_schema` _tool

`cdc_list_catalog_vocabulary` _tool

`cdc_query_dataset` _tool

`cdc_query_wonder` _tool

`cdc://datasets` _resource

`cdc://datasets/{datasetId}` _resource

`analyze_health_trend` _prompt