Medicare Part D MCP Server

15 actions. $200B+ in drug spending data. One tool call.

A Model Context Protocol (MCP) server giving AI assistants instant access to CMS Medicare Part D drug spending, prescriber, and pricing trend data — deployed on Cloudflare Workers with sub-50ms global response times.

Live endpoint: https://mcp-partd.medseal.app/mcp

No API key required. Public CMS data. Zero configuration.

What Can It Do?

Spending Analysis Get current quarterly spending or full multi-year trend analysis for any Part D drug. Covers 2024 Q1-Q4 quarterly data and 2019-2023 annual trends.

Drug Comparison Compare two drugs side-by-side: total spending, beneficiaries, cost per claim, year-over-year changes, and 4-year CAGR — all in one call.

Price Outlier Detection Surface drugs with the biggest price increases or decreases. Filter by risers, fallers, or both. Powered by CMS year-over-year change data.

Generic Alternative Finder Given any brand-name drug, find all drugs sharing the same generic compound and compare costs per beneficiary. Shows potential savings percentage.

Prescriber Intelligence Look up who prescribes what — by drug name, by state, or by NPI. Aggregate prescribing data with top drugs per provider.

Specialty Breakdown See which medical specialties prescribe a drug the most. Aggregated by provider type with claim counts and cost totals.

Geographic Patterns Prescribing data broken down by state — see where drugs are prescribed most heavily across the US.

Manufacturer Portfolio Look up any drug manufacturer to see their full Part D portfolio: all drugs, combined spending, beneficiary counts.

Dashboard Summary One-call overview of the entire Part D landscape: top drugs by spending, most prescribed by beneficiaries, biggest price movers.

Top Rankings Rank drugs by spending, beneficiaries, or claims. Sortable. Configurable result count.

Related MCP server: cms-datagov-mcp-server

Quick Start

Add to your Claude Desktop or Claude Code config:

{
  "mcpServers": {
    "partd": {
      "type": "url",
      "url": "https://mcp-partd.medseal.app/mcp"
    }
  }
}

That's it. Start asking about drug spending.

Actions

Examples

{"action": "summary"}
{"action": "compare", "drug": "Ozempic", "drug2": "Mounjaro"}
{"action": "outliers", "direction": "risers", "max_results": 10}
{"action": "generics", "drug": "Humira"}
{"action": "specialty", "drug": "Eliquis"}
{"action": "spending", "drug": "Ozempic"}
{"action": "prescribers", "drug": "Ozempic", "state": "CA"}
{"action": "geography", "drug": "Eliquis"}
{"action": "manufacturer", "manufacturer": "Eli Lilly"}
{"action": "top", "sort": "beneficiaries", "max_results": 20}
{"action": "search", "query": "insulin"}
{"action": "drug", "drug": "Eliquis", "dataset": "annual"}

Data Coverage

Source: CMS data.cms.gov — updated quarterly, no authentication required.

Target Users

• Health policy analysts — track drug spending trends, identify cost outliers

• Healthcare journalists — investigate pricing patterns, manufacturer portfolios

• Pharmacy benefit managers — compare drug costs, find generic alternatives

• Health IT developers — integrate Medicare spending data into applications

• AI/LLM builders — give your agents Medicare drug intelligence

• Researchers — prescribing patterns by specialty, geography, provider

Key Advantages

• No API key — public CMS data, zero configuration

• Token efficient — single tool with action dispatch, LLMs learn one schema

• Sub-50ms latency — Cloudflare Workers edge network, global deployment

• Analytical depth — not just lookups: comparisons, outlier detection, savings analysis

• 5 CMS datasets — spending, trends, prescribers, providers, geography

• Always current — CMS quarterly updates, 1-hour edge caching

Deploy Your Own

git clone https://github.com/stayce/partd-mcp-cloudflare
cd partd-mcp-cloudflare
npm install
npm run deploy

No secrets needed. No environment variables. Just deploy.

MedSeal MCP Suite

Part of a growing collection of healthcare MCP servers:

Together: diagnosis codes, functioning classification, drug spending, and prescriber data — all accessible to any MCP-compatible AI assistant.

License

MIT