CDC公共卫生数据查询MCP服务器

@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

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

Overview

CDC public health data from two systems: the Socrata-based CDC Open Data portal and 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

Resources

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

Prompts

Capability reference

cdc_discover_datasets _tool

• query, category, and tags filters (tags union, the other two intersect with them); up to 100 results per page (default 10; a larger limit, up to 10,000, is capped at 100 and the notice says so with the offset to continue from), with offset + limit capped at 10,000 (page_out_of_range past it); order is dataset_id (default, stable paging) or relevance
• Results carry assetType, columnCount (0 marks a chart, map, story, file, or link that yields no data), an 8-name columnSample, and a description cut to 300 characters (cdc_get_dataset_schema with full_description: true returns the whole text); enrichment adds totalCount and appliedFilters
• When a category or tags value matches nothing, the notice names the closest catalog values with their entry counts, e.g. Vaccination → "Vaccinations" (89 datasets)

cdc_get_dataset_schema _tool

• Four-by-four datasetId (e.g. bi63-dtpu); column_limit (default 100, any larger value up to the whole schema) and column_offset page wide schemas, reporting totalCount, truncated, and nextOffset
• The dataset description is capped at 4,000 characters by default, ending in … with descriptionTruncated and descriptionLength set when cut; full_description: true returns the whole text. Column descriptions are never shortened
• Columns carry fieldName, dataType, and description; rowCountSource says whether rowCount is a live count or Socrata's cached figure, which can lag an active dataset. A non-tabular asset fails with not_queryable

cdc_list_catalog_vocabulary _tool

• Every category comes back whole; the ~1,600 tags are ranked by entry count and paged with tag_limit (default 50; a value at or above the tag count returns them all) and tag_offset. filter (3+ characters) narrows both by word containment in either direction, not fuzzy matching
• Each value carries datasetCount; enrichment reports vocabularySize, categoryCount, tagCount, truncated, and nextOffset, with truncationCeiling bounding the count of every tag after the page — the tags before tag_offset rank at or above the first tag shown

cdc_query_dataset _tool

• select, where, group, having, order, and full-text search; up to 5,000 rows per call (default 100; a larger limit is capped at 5,000 and says so), offset up to 1,000,000. Pair offset paging with an order, set from the first page, ending in a per-row tie-breaker: :id on ungrouped rows, the grouped columns on a grouped query. When the order lacks one, the notice says to restart from offset 0 with it
• effectiveQuery echoes the SoQL clauses as sent, unencoded; truncated and nextOffset come from a one-row over-fetch, with no totalCount. A chart or map ID fails with not_queryable

cdc_query_wonder _tool

• group_by takes 1–4 of year, age_group, sex, race, national only. cause_icd10 and mcd_icd10 take an ICD-10 code, a chapter or block range (X40-X49), or a list of up to 50 matched as one union (an empty list reads as all causes). age_groups takes NS (age not stated) alone or with all eleven known groups, never beside only some of them, which CDC refuses. limit (any positive integer; the response budget bounds the page) and offset (max 10,000) page the table, and an empty page's notice says whether nothing matched or the offset ran past the end
• Rows carry deaths, population, crude_rate, and age_adjusted_rate where age can be standardized. Cells CDC returns as Suppressed, Unreliable, or Not Applicable read null and are listed in cellNotes; messages reports whole rows CDC hid, and totalCount is exact
• CDC rejects requests less than 15 seconds apart, so calls are spaced 16 seconds apart and queued one at a time

cdc://datasets _resource

• Top 50 catalog entries by popularity, each with id, name, assetType, category, columnCount, and updatedAt, plus the catalog totalCount
• columnCount: 0 marks a non-tabular entry; cdc_discover_datasets searches and pages the full catalog

cdc://datasets/{datasetId} _resource

• Dataset metadata plus the first 100 columns as application/json, with columnCount and truncated; the description is capped at 4,000 characters, with descriptionTruncated and descriptionLength set when cut
• Takes no selector; continue a wider schema with cdc_get_dataset_schema and column_offset, and read a cut description whole with its full_description: true. The notice names both when they apply

analyze_health_trend _prompt

• Arguments: topic required; timeRange and geography optional
• Returns one user message that routes the question to cdc_query_wonder (national mortality) or the Socrata tools (everything else), then walks discover → inspect → baseline → compare → synthesize

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:

• The Socrata SODA API v2.1 over the CDC Open Data portal (~1,080 datasets); no auth required, and an optional app token raises rate limits
• CDC WONDER mortality data from five databases covering 1999 through the current year, reached over its XML request API
• The three Socrata tools take a domain input, allowlisted to data.cdc.gov (default) and chronicdata.cdc.gov. Both front the same catalog, so it picks the host that answers, not the datasets you can reach
• category and tags match the catalog's controlled vocabularies (categories case-sensitively, tags case-insensitively), so a near miss returns nothing; cdc_list_catalog_vocabulary lists the real values
• Socrata returns every value as a string, and a column like year is numeric in some datasets and text in others. Match WHERE literals to the column's dataType from the schema: numbers bare, strings single-quoted

Agent-friendly output:

• Paging disclosed rather than implied: totalCount, truncated, nextOffset, and shown/cap. Both query tools bound each page by a 200,000-character budget over structuredContent and content[] together, so a wide page can end short of limit
• Typed error contracts with a recovery hint on every declared reason, such as not_queryable, page_out_of_range, and invalid_query
• Upstream gaps stay visible: WONDER status tokens per cell in cellNotes, hidden-row notices in messages, and rowCountSource on schema row counts
• effectiveQuery echoes the SoQL clauses or a WONDER query summary, so a result can be reproduced

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:

{
  "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.

{
  "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):

{
  "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:

{
  "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 (or Node.js v24+).
• Optional: a Socrata app token for higher rate limits.

Installation

1. Clone the repository:

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

2. Navigate into the directory:

cd cdc-health-mcp-server

3. Install dependencies:

bun install

4. Configure environment:

cp .env.example .env
# optionally set CDC_APP_TOKEN

Configuration

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

Running the server

Local development

• Build and run the production version:

  复制

  # One-time build
  bun run rebuild
  
  # Run the built server
  bun run start:http
  # or
  bun run start:stdio
  

• Run checks and tests:

  复制

  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

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 in src/index.ts
• 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.