MCP server

Give your coding agent the catalog. Pillwise speaks the Model Context Protocol, so an AI assistant can search medications and resolve codes while it builds your medication field.

The server is https://api.pillwise.dev/mcp (Streamable HTTP). It takes the same key as the API: a sandbox key is enough. Each tool call is one lookup on your plan, with the same limits.

Claude Code

claude mcp add --transport http pillwise https://api.pillwise.dev/mcp \
  --header "Authorization: Bearer pw_test_…"

Cursor

In .cursor/mcp.json:

{
  "mcpServers": {
    "pillwise": {
      "url": "https://api.pillwise.dev/mcp",
      "headers": {
        "Authorization": "Bearer pw_test_…"
      }
    }
  }
}

VS Code

In .vscode/mcp.json:

{
  "servers": {
    "pillwise": {
      "type": "http",
      "url": "https://api.pillwise.dev/mcp",
      "headers": {
        "Authorization": "Bearer pw_test_…"
      }
    }
  }
}

Other clients

Any client that supports remote MCP servers over HTTP with a custom header works: point it at https://api.pillwise.dev/mcp and send Authorization: Bearer with your key.

Tools

All of them read the catalog. None of them write anything, and none reach Cabinet or member data.

search_medications: Typeahead search by brand, ingredient or everyday name.
get_form_options: Form and strength choices for a search hit.
get_drug: The full record for a drug id.
get_packages: The package NDCs for a drug id.
lookup_ndc: The package and drug for an NDC, with or without hyphens.
lookup_rxcui: The clinical drug for an RxCUI.
list_recalls: Recalls for an NDC, a drug id, or since a date.
list_shortages: Shortages for an NDC or a drug id.
catalog_freshness: How current the catalog is.

Keep the key in your client's config, not in a prompt. Agents don't need a production key to build with: pw_test_ answers the same way.

API reference · OpenAPI spec