Skip to main content
Glama
Open Drug Data MCP
by knorby

Search NLM reference sources

drugdata_search_references
Read-onlyIdempotent

Find drug and medical references across PubMed, MeSH, RxClass, MedlinePlus, UMLS, and ClinicalTrials.gov by name or identifier, returning stable links and citation IDs.

Instructions

Search NLM reference sources beyond product and ingredient records: Prescribable RxNorm (the currently prescribable drug dataset), RxClass (drug classes by name or member RxCUI), MeSH (medical subject descriptors), PubMed (biomedical literature), MedlinePlus (consumer-health topic pages), MedlinePlus Genetics (condition and gene pages), MedlinePlus Connect (drug pages by name, RxCUI, or NDC), UMLS (unified medical terminology; needs UMLS_API_KEY), and ClinicalTrials.gov (study registry). One call searches exactly one source — pass source plus the query. Results are link-first references: a title, the stable identifier (RxCUI, class id, MeSH id, PMID, CUI, or NCT number), and a deep link to the upstream record — not extracted article or study text. Each source accepts only some query types (for example MeSH searches by name only); an invalid pairing is rejected before any request. Continue with the returned cursor when hasMore is true. No matches is a normal empty result, not an error.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "source": "<source>",
      "query": "<query>"
    },
    "name": "drugdata_search_references"
  }
}

Input Schema

NameRequiredDescriptionDefault
typeNoWhat the query holds: 'name' (default), 'rxcui', 'ndc', or 'cui'. Not every source accepts every type; an invalid pairing is rejected with the types that source accepts.name
limitNoMaximum references to return. Sources that page client-side may return slightly fewer when the upstream list ends early.
queryYesThe search value: a name or term by default, or the identifier named by `type`.
cursorNoContinuation token from a previous result's nextCursor. Only valid for the same source, type, query, and limit.
sourceYesWhich reference source to search: 'prescribable', 'rxclass', 'mesh', 'pubmed', 'medlineplus', 'medlineplus_genetics', 'medlineplus_connect', 'umls', or 'clinicaltrials'.
response_formatNoText format: 'markdown' for a readable reference list, or 'json' for the structured result serialized as JSON.markdown

Output Schema

NameRequiredDescriptionDefault
typeYes
queryYes
groupsYes
sourceYes
nextStepsYes
truncatedYesWhether bounding shortened any content.

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv0.1.0

TDQS

A4.6/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Beyond the annotations (readOnly, idempotent, openWorld, non-destructive), the description discloses the return shape (link-first references with title, stable identifier, and deep link), explicitly warns it is "not extracted article or study text," documents pre-flight validation of invalid source/type pairings, describes pagination via `cursor`/`hasMore`, and flags that UMLS requires UMLS_API_KEY. That is meaningful operational context an agent would otherwise have to discover by failing a call.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with the source catalog and then the critical constraints (one source per call, link-first results, cursor continuation). The long comma-separated source enumeration makes it dense, but each clause carries real information; only the repeated source list (which duplicates the `source` enum) borders on redundancy.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a six-parameter, nine-source federated search tool with an output schema present, the description covers the decisions an agent must make: which source, which query type it accepts, how to page, and what the results contain. Nothing material for correct invocation is left unstated.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so baseline is 3, but the description genuinely adds meaning: it explains that only one source may be searched per call, that each source accepts only a subset of query types (with MeSH as a name-only example), and that the invalid pairing is rejected with the accepted types. It does not add much on `limit`, `response_format`, or cursor scoping beyond what the schema already says.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Opens with a specific verb+resource ("Search NLM reference sources") and enumerates all nine supported sources with a parenthetical gloss for each. It explicitly scopes itself against the sibling tools by saying these are references "beyond product and ingredient records," so an agent can separate it from drugdata_search_products/ingredients without opening a schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

States the core usage contract clearly: "One call searches exactly one source — pass `source` plus the query," and notes that source/query-type pairings are validated with rejection before any request. It also says to continue with `cursor` when `hasMore` is true and that an empty result is normal, not an error. What it lacks is explicit routing language (e.g., "for product records use drugdata_search_products instead"), leaving the alternative-tool decision to inference from "beyond product and ingredient records."

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.