Pubmed Mcp Server

---

Overview

The biomedical literature via NCBI's E-utilities, PubMed Central, and Europe PMC. Search it, fetch metadata and full text, resolve identifiers and partial citations, format references, and ground queries in MeSH vocabulary. Runs as a stdio process, a local Streamable HTTP server, or the public hosted endpoint above.

Tools

Tool	Description
pubmed_search_articles	Search PubMed with full query syntax, field-specific filters, date ranges, pagination, and optional brief summaries
pubmed_europepmc_search	Search Europe PMC for preprints, patents, Agricola, and EPMC-only OA records that don't surface in PubMed. Cursor-based pagination.
pubmed_europepmc_fetch	Fetch complete Europe PMC records — including the untruncated abstract — by source + epmcId, the only identifier many preprint, patent, and Agricola records carry
pubmed_fetch_articles	Fetch full article metadata by PMIDs — abstract, authors, journal, MeSH terms, grants
pubmed_fetch_fulltext	Fetch full-text articles via a chain: NCBI PMC EFetch → Europe PMC fullTextXML → Unpaywall. Accepts PMIDs, PMCIDs, or DOIs.
pubmed_format_citations	Generate formatted citations in APA 7th, MLA 9th, BibTeX, RIS, or Vancouver (ICMJE/NLM)
pubmed_find_related	Find similar articles, citing articles, or references for a given PMID
pubmed_spell_check	Spell-check a biomedical query via NCBI ESpell — returns the corrected query and whether a suggestion was found
pubmed_lookup_mesh	Search MeSH by heading — tree numbers, scope notes, entry terms — for building controlled-vocabulary queries
pubmed_lookup_citation	Resolve partial bibliographic references to PubMed IDs via ECitMatch
pubmed_convert_ids	Convert between DOI, PMID, and PMCID using the PMC ID Converter API

Resources

Resource	Description
pubmed://database/info	PubMed database metadata via EInfo (field list, record count, last update)

Prompts

Prompt	Description
research_plan	Generate a structured 4-phase biomedical research plan outline

Capability reference

pubmed_search_articles _tool

• Full PubMed boolean and field-tag syntax, plus structured filters: author, journal, MeSH terms, language, species, publication type, has-abstract, free-full-text
• Date ranges by publication, modification, or Entrez date; sort by relevance, date, author, or journal; offset pagination
• Optional brief summaries for the top N results via ESummary
• NCBI Bookshelf hits carry bookTitle, publisherName, docType, and editors in place of the empty source
• Echoes the original query, the fully applied PubMed query, and normalized filter metadata

---

pubmed_fetch_articles _tool

• Up to 200 PMIDs per call (POST for batches of 100 or more)
• Title, abstract, authors with deduplicated affiliations, journal info, DOI, PubMed/PMC links; optional MeSH terms, grants, and publication types
• Tolerant of PubMed's inconsistent XML — structured abstracts, missing fields, varying date formats
• Bookshelf chapters and books are first-class: recordType (journal-article / book-chapter / book) plus a book object (title, publisher, editors, ISBNs, Bookshelf accession); journalInfo is absent on them
• Article-number journals report journalInfo.elocationId + elocationIdType rather than a page range
• Opt-in maxResponseCharacters keeps whole records in order until the ceiling, then defers the rest to deferred.ids for a follow-up call

---

pubmed_fetch_fulltext _tool

• Exactly one of pmcids, pmids, or dois (one id per element), up to 10 per request
• Three-tier chain: NCBI PMC EFetch → Europe PMC fullTextXML (EUROPEPMC_ENABLED, default on) → Unpaywall (needs UNPAYWALL_EMAIL); viaSource names which tier served each article
• Preprints, patents, and Agricola records have metadata via pubmed_europepmc_search but no full text through this chain — Europe PMC's fullTextXML is PMC-keyed
• source: "pmc" returns structured sections plus tables[] (cells, caption, label, footnotes) and assets[] (figures and supplementary material, with [Figure: <label>] markers left in the body); source: "unpaywall" returns a best-effort body with contentFormat (html-markdown / pdf-text)
• Unavailable entries carry a typed reason (not-found, no-doi, doi-lookup-failed, no-oa, service-error, …), idType, triedTiers (per-tier outcome in execution order), and unqueriedTiers when an unconfigured tier could have served the id
• Filters and budgets: sections (case-insensitive title match), maxSections, includeTables, includeAssets, maxCharacters, maxCharactersPerSection, overflowMode (truncate / outline), and maxResponseCharacters, which defers whole articles past the ceiling to deferred.ids; a truncation object reports what was shortened or omitted

---

pubmed_europepmc_search _tool

• Reaches records PubMed can't: preprints (PPR), patents (PAT), Agricola (AGR), alongside MED and PMC; default sources is ["MED", "PMC", "PPR"]
• Cursor pagination via cursorMark — * for the first page, then nextCursorMark
• Hits carry source plus pmid / pmcId / doi when known; abstractSnippet is capped at 400 characters, with abstractTruncated flagging the cut
• Not registered when EUROPEPMC_ENABLED=false

---

pubmed_europepmc_fetch _tool

• Full records with the untruncated plain-text abstract, addressed by source + epmcId — the only identifier preprint, patent, and Agricola records reliably carry
• Up to 25 per call in one Europe PMC request; unresolved ids come back in notFound rather than failing the batch
• Not registered when EUROPEPMC_ENABLED=false

---

pubmed_format_citations _tool

• APA 7th, MLA 9th, BibTeX, RIS, Vancouver (ICMJE/NLM); several styles per article in one call, up to 50 articles
• Bookshelf chapters and books cite in each style's edited-book form; articles without a page range cite by electronic locator in each style's convention
• Hand-rolled formatters — zero dependencies, Workers-compatible
• Reports formatted counts and unavailable PMIDs

---

pubmed_find_related _tool

• similar, cited_by, or references for a PMID, in NCBI relevance order, enriched with title, authors, date, and source (or Bookshelf book title and publisher)
• Falls back to Europe PMC, then OpenAlex, when NCBI can't answer; the response names the provider. Fails with a typed all_providers_failed error rather than an empty result

---

pubmed_spell_check _tool

• Runs a query through NCBI ESpell and returns original, corrected, and hasSuggestion
• A blank or whitespace-only query is rejected rather than sent upstream

---

pubmed_lookup_mesh _tool

• Looks up MeSH descriptors by name or free-text term, pinning the exact-heading match to the top of the first page
• Records carry meshId (DescriptorUI), entrezUid, and, with includeDetails (default on), tree numbers, scope notes, and entry terms
• maxResults up to 50 with offset pagination via nextOffset; totalCount reports the upstream match count

---

pubmed_lookup_citation _tool

• Match on journal, year, volume, first page, and/or author — at least one field, more fields for better precision; up to 25 per call
• Pipes and line breaks are rejected at the schema (ECitMatch's wire format is pipe-delimited); the free-form key label is exempt
• Explicit matched, not_found, and ambiguous statuses with recovery detail

---

pubmed_convert_ids _tool

• Up to 50 DOIs, PMIDs, or PMCIDs per call, all one type; only PMC-indexed articles resolve
• One id per element — a packed "23193287,37952131" is rejected rather than expanded
• Per-id success/error rows; a partial batch never fails as a whole

---

pubmed://database/info _resource

• Live EInfo call for the pubmed database, returned as application/json
• dbName, description, count, lastUpdate, and fields[] — each field's short name (the tag usable in pubmed_search_articles queries), fullName, and description
• No parameters

---

research_plan _prompt

• Arguments: title, goal, keywords (comma-separated) required; organism and includeAgentPrompts ("true" / "false") optional
• Returns two messages: an assistant framing message (biomedical research planning assistant, grounds recommendations in the PubMed tools when available) and a user message carrying the plan
• The plan walks four phases — Conception & Planning, Data Collection & Processing, Analysis & Interpretation, Dissemination — with sub-steps under each
• includeAgentPrompts: "true" adds an agent-guidance block under each sub-step, several of which point at pubmed_search_articles and pubmed_lookup_mesh

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.

PubMed-specific:

• Complete NCBI E-utilities integration (ESearch, EFetch, ESummary, ELink, ESpell, EInfo, ECitMatch) plus PMC ID Converter
• Sequential request queue with configurable delay for NCBI rate limit compliance
• NCBI-specific XML parser with isArray hints for PubMed's inconsistent XML structure
• Hand-rolled citation formatters (APA, MLA, BibTeX, RIS, Vancouver) — zero deps, Workers-compatible

Agent-friendly output:

• Provenance on every response — source labels, license fields, best-effort warnings on Unpaywall results, and effective-query echo on searches so agents can reason about trust
• Graceful partial failure — batch tools return per-item success/error rows instead of failing the request, with structured status codes and actionable next-step text
• Discriminated output contracts — source: "pmc" | "unpaywall", typed unavailable reasons, viaSource and triedTiers fields — callers branch on data, not string parsing

Getting started

Public Hosted Instance

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

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

Self-Hosted / Local

Add the following to your MCP client configuration file.

{
  "mcpServers": {
    "pubmed-mcp-server": {
      "type": "stdio",
      "command": "bunx",
      "args": ["@cyanheads/pubmed-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_LOG_LEVEL": "info",
        "NCBI_API_KEY": "your-key-here"
      }
    }
  }
}

Or with npx (no Bun required):

{
  "mcpServers": {
    "pubmed-mcp-server": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@cyanheads/pubmed-mcp-server@latest"],
      "env": {
        "MCP_TRANSPORT_TYPE": "stdio",
        "MCP_LOG_LEVEL": "info",
        "NCBI_API_KEY": "your-key-here"
      }
    }
  }
}

Or with Docker:

{
  "mcpServers": {
    "pubmed-mcp-server": {
      "type": "stdio",
      "command": "docker",
      "args": ["run", "-i", "--rm", "-e", "MCP_TRANSPORT_TYPE=stdio", "ghcr.io/cyanheads/pubmed-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: NCBI API key for higher rate limits (10 req/s vs 3 req/s).

Installation

1. Clone the repository:

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

2. Navigate into the directory:

cd pubmed-mcp-server

3. Install dependencies:

bun install

Configuration

Key environment variables:

Variable	Description	Default
MCP_TRANSPORT_TYPE	Transport: stdio or http	stdio
MCP_HTTP_PORT	HTTP server port	3010
MCP_HTTP_ENDPOINT_PATH	HTTP endpoint path where the MCP server is mounted	/mcp
MCP_SESSION_MODE	HTTP session mode: stateless, stateful, or auto (resolves to stateful). This server ships stateless — it has no ctx.requestInput call sites.	stateless
MCP_PUBLIC_URL	Public origin override for TLS-terminating reverse-proxy deployments (landing page, Server Card, RFC 9728 metadata).	none
MCP_AUTH_MODE	Authentication: none, jwt, or oauth	none
MCP_LOG_LEVEL	Log level (debug, info, warning, error, etc.)	info
MCP_GC_PRESSURE_INTERVAL_MS	Opt-in Bun-only forced-GC pressure loop (ms). Drains the per-request McpServer/McpSessionTransport cycle under sustained low-traffic HTTP. Recommended starting point if heap growth is observed: 60000.	0 (disabled)
LOGS_DIR	Directory for log files (Node.js only). Relative paths resolve against the application root.	<app-root>/logs
STORAGE_PROVIDER_TYPE	Storage backend: in-memory, filesystem, supabase, cloudflare-kv/r2/d1	in-memory
NCBI_API_KEY	NCBI API key for higher rate limits (10 req/s vs 3 req/s)	none
NCBI_ADMIN_EMAIL	Contact email sent with NCBI requests (recommended by NCBI)	none
NCBI_REQUEST_DELAY_MS	Minimum gap between NCBI request starts in ms	334 (100 with key)
NCBI_MAX_CONCURRENT	Max concurrent in-flight NCBI requests	8
NCBI_MAX_RETRIES	Retry attempts for failed NCBI requests	6
NCBI_TIMEOUT_MS	Per-request HTTP timeout in ms	30000
NCBI_TOTAL_DEADLINE_MS	Total deadline across all retry attempts for one NCBI call, in ms	60000
UNPAYWALL_EMAIL	Contact email for Unpaywall. When set, pubmed_fetch_fulltext falls back to Unpaywall open-access copies for non-PMC DOIs	none
UNPAYWALL_TIMEOUT_MS	Per-request HTTP timeout for Unpaywall lookups and content fetches, in ms	20000
EUROPEPMC_ENABLED	Enable Europe PMC search tool and the pubmed_fetch_fulltext JATS fallback chain. Set false to disable all EPMC calls and skip tool registration.	true
EUROPEPMC_EMAIL	Optional contact email sent with Europe PMC requests (EBI courtesy).	none
EUROPEPMC_REQUEST_DELAY_MS	Minimum gap between Europe PMC request starts in ms	200
EUROPEPMC_MAX_RETRIES	Retry attempts for failed Europe PMC requests	3
EUROPEPMC_TIMEOUT_MS	Per-request HTTP timeout for Europe PMC calls, in ms	20000
OTEL_ENABLED	Enable OpenTelemetry	false

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
  

Project structure

Directory	Purpose
src/mcp-server/tools	Tool definitions (*.tool.ts). Eleven tools across PubMed, PMC, and Europe PMC.
src/mcp-server/resources	Resource definitions. Database info resource.
src/mcp-server/prompts	Prompt definitions. Research plan prompt.
src/services/ncbi	NCBI E-utilities service layer — API client, queue, parser, formatter.
src/services/europe-pmc	Europe PMC service — search + fullTextXML JATS retrieval. Reuses the NCBI JATS parser.
src/services/unpaywall	Unpaywall service — DOI → OA location resolution and content fetch (HTML/PDF).
src/config	Server-specific environment variable parsing and validation with Zod.
tests/	Unit and integration tests, mirroring the src/ 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

Contributing

Issues are welcome. Run checks and tests before submitting:

bun run devcheck
bun run test

License

This project is licensed under the Apache 2.0 License. See the LICENSE file for details.