Noodle Biomedical Literature Discovery MCP

What Noodle Biomedical Literature Discovery MCP does

Noodle Biomedical Literature Discovery MCP MCP server exposes a public, read-only literature discovery service from Helena Bioinformatics. It is intended for finding and inspecting biomedical publications rather than modifying records or submitting private material. The searchable collection is derived from PubMed and can be queried with natural-language topics or identifiers such as PMID, DOI, and PMCID.

The service supports two kinds of work: retrieving information about known publications and discovering additional papers. An agent can resolve an identifier, inspect publication metadata, identify related works, and follow bounded citation or semantic relationships. Corpus-level metadata is also available for checking sources, freshness, size, coverage, and active graph information.

How it works

Clients connect to the hosted Streamable HTTP endpoint at https://api.helena.bio/noodle/v1/mcp. Any MCP client with remote Streamable HTTP support can use the service. No local npm or npx package is required, and the public service does not require or accept an account, API key, patient data, or private content.

A typical exploration begins with a PMID or Noodle work ID. The client requests one bounded neighborhood at a time, retains the returned work identifiers, and can continue from those identifiers for further exploration. Results preserve edge types and graph provenance. Search ranking, citation distance, semantic similarity, co-mention, and graph distance are discovery signals; they are not evidence of causality, scientific validity, diagnosis, or treatment effectiveness.

Tools and capabilities

The Noodle Biomedical Literature Discovery MCP MCP server publishes seven actions:

• search_biomedical_literature searches the public biomedical corpus.
• get_publication_details retrieves source-linked details by PMID.
• get_work_details retrieves details by Noodle work ID.
• get_publication_neighborhood explores bounded relationships from a publication.
• get_work_neighborhood explores bounded relationships from a work.
• get_corpus_summary reports corpus and graph metadata.
• support_helena is a separate, explicit opt-in information action.

These capabilities support workflows such as finding papers on a gene or disease, resolving a citation identifier, comparing nearby literature, and walking a constrained graph while retaining relationship types. The repository also includes guidance and integrations for platforms including Dify, n8n, Galaxy, KNIME, notebooks, and Cytoscape, but the core service is accessed through MCP.

Limitations and notes

Noodle Biomedical Literature Discovery MCP MCP server is read-only. It should not be used to send patient information, clinical records, credentials, private case details, or private uploaded content to the public endpoint or issue tracker. The graph must be traversed in bounded steps, with a visited-ID set and a stop condition when no neighborhood is returned.

The service helps discover and organize research; it does not establish that a paper is scientifically valid, that two works have a causal relationship, or that a finding supports a diagnosis or treatment. Developers building research agents should present the returned source links and distinguish discovery signals from conclusions. The project is distributed under the Apache License 2.0.