Skip to content

Latest commit

 

History

History
315 lines (230 loc) · 14.5 KB

File metadata and controls

315 lines (230 loc) · 14.5 KB

Connect: External MCP Servers for Life Sciences

Add literature search, clinical trials, preprint access, and ML model discovery to your precision medicine workflow. These run alongside our custom servers — no changes to existing platform configuration.

graph LR
    subgraph hosted[Anthropic Connectors]
        CT[ClinicalTrials.gov<br/>6 tools]
        BIO[bioRxiv & medRxiv<br/>9 tools]
        PM[PubMed<br/>5 tools]
        SQ[Seqera<br/>7 tools]
    end
    subgraph local[Community Servers]
        CB[cBioPortal · TCGA<br/>12 tools]
        HF[Hugging Face<br/>7 tools]
    end
    style hosted fill:#f0f8f0,stroke:#4a9
    style local fill:#f0f0f8,stroke:#49a
    style CT fill:#e8f4e8,stroke:#4a9
    style BIO fill:#e8f4e8,stroke:#4a9
    style PM fill:#e8f4e8,stroke:#4a9
    style SQ fill:#e8f4e8,stroke:#4a9
    style CB fill:#e8eaf4,stroke:#49a
    style HF fill:#e8eaf4,stroke:#49a
Loading
Anthropic Connectors (hosted) Community MCP Servers (self-hosted)
What Remote servers hosted by Anthropic Open source servers you run locally
Setup Toggle on in Settings > Connectors Add config JSON or use claude mcp add CLI
Install Nothing to install Requires Node.js, Python, or similar
Auth Handled by Anthropic You manage API keys and tokens
Examples below ClinicalTrials.gov, bioRxiv, PubMed, Seqera cBioPortal (TCGA), Hugging Face

Anthropic Connectors (hosted)

Remote MCP servers hosted by Anthropic. Enable with a toggle — no install, no config files. Anthropic manages the infrastructure.

Setup (same for all connectors): Claude Desktop > Settings > Connectors > find the connector > click Connect

Auth note: ClinicalTrials.gov, bioRxiv, and PubMed require no account or API key. Seqera requires you to authenticate with a free Seqera Platform account (cloud.seqera.io — free Cloud Basic tier, no credit card) during the connector setup flow. No env vars or API tokens to manage — the connector handles auth after you sign in.

ClinicalTrials.gov

Search 500,000+ FDA-regulated clinical studies. Find trials by condition, sponsor, phase, or patient eligibility. Analyze endpoints across trials.

Available Tools (6)

Tool What it does Example use
search_trials Search by condition, intervention, location, status, phase Find recruiting ovarian cancer trials
get_trial_details Full protocol for a specific NCT ID Get eligibility criteria for NCT04567890
search_by_sponsor All trials by a company or institution Analyze AstraZeneca's PARP inhibitor pipeline
search_investigators Find PIs and research sites Identify ovarian cancer investigators at Dana-Farber
analyze_endpoints Compare outcome measures across trials Aggregate Phase 3 ovarian cancer endpoints
search_by_eligibility Match patient criteria to open trials Find trials for 65yo female with BRCA1+ HGSOC

Example Prompts

"Search ClinicalTrials.gov for recruiting Phase 2/3 trials testing PARP inhibitors
in platinum-resistant ovarian cancer."

"Find all trials sponsored by AstraZeneca for ovarian cancer and summarize their
primary endpoints by phase."

"What recruiting trials could a 58-year-old female with BRCA1-mutated, Stage IV
HGSOC, ECOG 1, who has failed two prior lines of platinum therapy, be eligible for?"

bioRxiv & medRxiv

Access 260,000+ preprints across 27 bioRxiv and 40+ medRxiv categories. Find the latest research before peer review, track which preprints have been published, and analyze submission trends.

Available Tools (9)

Tool What it does Example use
search_preprints Search by date range, subject, or recent submissions Find spatial transcriptomics preprints from last month
get_preprint Full metadata, abstract, author info, publication status Get details for a specific preprint DOI
get_categories List all 27 bioRxiv subject categories Browse available disciplines
search_published_articles Track preprints that became peer-reviewed papers Check if a preprint was published in a journal
search_biorxiv_publications Simplified bioRxiv publication tracking Quick publication status lookup
search_publisher_articles Filter by specific journals or publishers Find preprints published in Nature Cancer
search_by_funder Discover research by funding organization Find NCI-funded ovarian cancer preprints
get_content_statistics Platform submission volumes and growth Analyze preprint trends in oncology
get_usage_statistics Abstract views, full-text views, PDF downloads Gauge research impact before publication

Example Prompts

"Search bioRxiv for preprints on tumor microenvironment spatial transcriptomics
in ovarian cancer from the last 6 months."

"Find recent medRxiv preprints about PARP inhibitor combinations and platinum
resistance mechanisms."

"Look up preprints on single-cell RNA sequencing of high-grade serous ovarian
cancer and check which ones have been published in peer-reviewed journals."

PubMed

Search 36M+ biomedical citations from the NLM. Find articles by keyword, author, or journal. Retrieve abstracts, full text (when available via PMC), and related articles. Convert between PMID, PMC ID, and DOI formats.

Capabilities

Capability Example use
Article search Find ovarian cancer immunotherapy papers by keyword, author, or MeSH term
Metadata retrieval Get abstracts, authors, publication dates, citations for a PMID
Full-text access Retrieve complete articles from PubMed Central (when available)
Related articles Discover similar research across NCBI databases
Citation matching Convert between PMID, PMC ID, and DOI formats

Example Prompts

"Search PubMed for recent papers on PARP inhibitor resistance in BRCA1-mutated
ovarian cancer, then summarize the top 3 findings."

"Find publications about PIK3CA H1047R mutation and targeted therapy response
in high-grade serous ovarian cancer."

"Search for spatial transcriptomics studies of the tumor microenvironment in
ovarian cancer published since 2023."

Seqera

Nextflow workflow orchestration via Seqera Platform. Launch nf-core pipelines, monitor runs, search modules, and troubleshoot failures via natural language. Free for all Seqera users.

Cost: The Seqera connector itself is free. Seqera Platform has a free Cloud Basic tier (250 runs, 3 users). Academic institutions can apply for free Cloud Pro. Compute (if using Seqera Cloud): $0.10/CPU-hour, $0.025/GiB-hour. See seqera.io/pricing.

Available Tools (7)

Tool What it does Example use
call_seqera_api Call Seqera Platform API endpoints Get details on a specific workflow run
search_seqera_api Search across Seqera Platform resources Find recent failed runs in your workspace
call_data_tool Manage and access datasets in Seqera List datasets available for a pipeline
search_data_tool Search for data tools and resources Find data tools for spatial transcriptomics
describe_nfcore_module Get details on a specific nf-core module Describe the STAR alignment module
search_nfcore_module Search the nf-core module registry Find modules for variant calling
nfcore_suggest_analysis Suggest nf-core pipelines for your analysis Recommend a pipeline for RNA-seq of HGSOC samples

Example Prompts

"What nf-core pipelines are available for spatial transcriptomics analysis?"

"Search Seqera for my recent workflow runs and tell me if any failed.
If so, what went wrong?"

"Suggest an analysis pipeline for bulk RNA-seq of platinum-resistant
high-grade serous ovarian cancer samples, and describe the key modules."

Community MCP Servers (self-hosted)

Open source servers you install and run locally. Configure via claude_desktop_config.json or claude mcp add CLI. The server process runs on your machine (or connects to a remote endpoint).

cBioPortal (TCGA & Cancer Genomics)

Query cancer genomics data from cBioPortal — the open-access platform hosting all 33 TCGA cancer types plus 200+ additional studies. Search mutations, gene expression, copy number, clinical data, and molecular profiles. No authentication required for the public portal.

Why cBioPortal for TCGA? Research hospitals access TCGA data two ways: ~50% query the public databases, ~50% use on-premise snapshots loaded into local cBioPortal instances. This server supports both — point it at the public portal or your hospital's instance via CBIOPORTAL_BASE_URL.

Cost: Free. The public cBioPortal API has no authentication, no rate limits, and no usage fees. Local instances require your own infrastructure.

Setup

Claude Code CLI:

git clone https://github.com/pickleton89/cbioportal-mcp.git
cd cbioportal-mcp
uv sync

Then add to Claude Code:

claude mcp add cbioportal-mcp -s local \
  -- uv run cbioportal-mcp

For a hospital's local cBioPortal instance, set the base URL:

export CBIOPORTAL_BASE_URL=https://cbioportal.yourhospital.org/api

Claude Desktop: Add to claude_desktop_config.json:

{
  "mcpServers": {
    "cbioportal": {
      "command": "uv",
      "args": ["run", "cbioportal-mcp"],
      "cwd": "/path/to/cbioportal-mcp"
    }
  }
}

Available Tools (12)

Tool What it does Example use
get_cancer_studies List available cancer studies with pagination Browse all TCGA cohorts
search_studies Keyword-based study discovery Find ovarian cancer studies
get_study_details Comprehensive study metadata Get details on TCGA-OV (ovarian)
get_samples_in_study Sample data within a study List all TCGA-OV tumor samples
get_genes Gene information by ID or symbol Look up TP53, BRCA1, PIK3CA
search_genes Keyword search for genes Find genes in the PI3K pathway
get_mutations_in_gene Mutation details within a study Get TP53 mutations across TCGA-OV
get_clinical_data Patient clinical information Retrieve survival, stage, treatment data
get_molecular_profiles Available molecular data types per study List expression, mutation, CNV profiles
get_gene_panels_for_study Gene panel information Check which genes were sequenced
get_multiple_studies Concurrent multi-study fetch Compare TCGA-OV, TCGA-BRCA, TCGA-LUAD
get_multiple_genes Concurrent batched gene retrieval Batch lookup of DNA repair pathway genes

Example Prompts

"Search cBioPortal for TCGA ovarian cancer studies and show me TP53 mutation
frequency compared to other TCGA cohorts."

"Get clinical data for TCGA-OV including survival outcomes and platinum
sensitivity status."

"What are the most frequently mutated genes in TCGA high-grade serous
ovarian cancer samples? Compare to TCGA breast cancer."

"Look up PIK3CA, PTEN, and AKT1 mutations in TCGA-OV and show the
mutation types (missense, nonsense, frameshift)."

Hugging Face

Search and explore ML models, datasets, Spaces, and papers on the Hugging Face Hub. Run community tools via MCP-compatible Gradio apps. Useful for finding genomic foundation models, biomedical NLP models, and relevant datasets.

Setup

Claude Desktop: Add via Settings > Connectors > find "Hugging Face" > click Connect. Or configure manually — visit huggingface.co/settings/mcp, select your client, and copy the config snippet.

Claude Code CLI:

claude mcp add hf-mcp-server -t http https://huggingface.co/mcp?login

Then start claude and follow the authentication prompt. Or use an explicit token:

claude mcp add hf-mcp-server \
  -t http https://huggingface.co/mcp \
  -H "Authorization: Bearer <YOUR_HF_TOKEN>"

Get a free HF token: huggingface.co/settings/tokens

Built-in Tools (7)

Tool What it does Example use
Spaces Semantic Search Find AI apps via natural language Find a tool that can segment cell images
Papers Semantic Search Find ML research papers Search for papers on genomic foundation models
Model Search Search models with filters for task, library Find biomedical text classification models
Dataset Search Search datasets with filters for author, tags Find single-cell RNA-seq datasets
Documentation Semantic Search Search HF docs using natural language How do I fine-tune with PEFT?
Run and Manage Jobs Run and monitor jobs on HF infrastructure Schedule a model training job
Hub Repository Details Get details about models, datasets, Spaces Get info on a specific model card

You can also add community Gradio Spaces as extra tools at huggingface.co/settings/mcp.

Example Prompts

"Search Hugging Face for biomedical language models fine-tuned on PubMed data."

"Find datasets related to ovarian cancer gene expression on Hugging Face."

"Search Hugging Face papers for recent work on spatial transcriptomics
foundation models."

References

Anthropic connectors:

Community servers: