Noodle Biomedical Literature Discovery MCP is the official public, read-only Model Context Protocol (MCP) adapter from Helena Bioinformatics, focused on biomedical literature discovery. Its public endpoint is https://api.helena.bio/noodle/v1/mcp , and its official registry identifier is io.github.helena-bioinformatics/noodle . Using this service requires no account, API key, patient data, or private content, and it does not accept these either. Any MCP client that supports remote Streamable HTTP can use this endpoint.
Main capabilities The agent can perform the following operations:
- Search the literature in a public PubMed-derived biomedical corpus via natural language, PMID, DOI, or PMCID;
- Retrieve publication records with source links via PMID or Noodle work ID;
- Starting from a publication, traverse bounded citation and semantic neighborhoods;
- Use the returned work identifiers to continue graph exploration while preserving edge types and graph provenance information;
- Inspect corpus size, sources, freshness, coverage, and active graph metadata.
The service publishes seven tools: search_biomedical_literature, get_publication_details, get_work_details, get_publication_neighborhood, get_work_neighborhood, get_corpus_summary, and a separate, explicitly opt-in support_helena informational operation.
How to use Any MCP client that supports remote Streamable HTTP can connect to this endpoint. Specific configuration methods for ChatGPT, Claude, Codex, VS Code, Cursor, Windsurf, Gemini CLI, Grok, Perplexity, Microsoft Copilot Studio, Biomni, and Biorouter are located in the repository's registry/platforms and integrations directories.
The repository also includes ready-to-use packages for the following ecosystems:
- Dify, including a reproducible .difypkg;
- n8n, using the exact stateless MCP JSON-RPC contract supported by the hosted service;
- Galaxy, with a Planemo linted ToolShed wrapper;
- KNIME Analytics Platform, with a table-to-MCP Python Script node and a prepared Hub list;
- Google Colab and Kaggle notebooks, as well as a Cytoscape GraphML workflow;
- A companion Galaxy Training Network tutorial for the Folklore-to-Noodle literature workflow.
The companion Agent Skill is located at skills/noodle-biomedical-literature-discovery . It supports implicit, task-first selection, for example it can handle the following requests:
- "Find source-linked papers about BRCA1 homologous recombination."
- "What publication is PMID 35008774?"
- "Show papers related to this article via citations and semantic similarity."
- "Walk two hops from this work ID and preserve edge types."
Commands for building a deterministic skill archive and for running a brand-blind contract audit can be found in the repository instructions. The benchmark contains 60 prompts, omits words such as Noodle, Helena, and MCP, and covers all six scientific paths as well as negative and safety controls.
Use cases and limitations This tool is suitable for scenarios that require rapid discovery of relevant papers from the biomedical literature, retrieval of specific numbered publications, and exploration of citation and semantic association networks, and is especially suitable as a literature discovery aid in conversational AI clients. When using it, start from a resolved PMID or work ID and request one bounded neighborhood at a time. Returned edges must be reported truthfully, a set of visited IDs must be maintained, and you must stop when a missing neighborhood is encountered. Search ranking, citation proximity, semantic similarity, co-occurrence mentions, and graph distance are only discovery signals and must not be used to establish causality, scientific validity, diagnosis, or treatment basis.
Other information Python 3.12 is the required environment. This project is licensed under the Apache License 2.0. The privacy policy is at https://noodle.helena.bio/privacy The persistent research resource identifier is RRID:SCR_028920, and in the Methods section it may be cited as Noodle (RRID:SCR_028920). The support contact is https://noodle.helena.bio/contact or contact@helena.bio. Public resources include practical tutorials, connector and agent selection guides, client integrations, Server Card, official registration, citable versions, Software Heritage archive, and methodology pages. For security vulnerabilities, please report privately according to SECURITY.md, and do not submit patients, private cases, clinical records, credentials, or private uploads to public services or issue trackers.
