Folklore Clinical Variant Interpretation MCP

What Folklore Clinical Variant Interpretation MCP MCP server does

The Folklore Clinical Variant Interpretation MCP MCP server exposes public evidence for the interpretation stage of human genomic analysis. It is intended for cases where a variant has already been identified and represented using supported HGVS, rsID, or verified GRCh38 allele information. It can return resolved variant identity, automated classification, interpretation criteria, available evidence, source versions, provenance, and stated uncertainty.

The server also covers gene-disease and disease-gene research. A gene symbol such as BRCA1 or an HGNC identifier can be used to find associated diseases. A disease name or exact MONDO identifier can be used to search for related genes. Results preserve inheritance information, source classifications, and provenance where available. Literature retrieval supports investigation of publications linked to a variant or biological entity; a literature association does not by itself establish pathogenicity.

How it works

Requests are sent to the hosted MCP endpoint:

https://api.helena.bio/folklore/v1/mcp

The service uses stateless Streamable HTTP and supports MCP operations including server discovery, tool listing, tool calls, resource listing, and resource reading. The hosted SDK also accepts the documented legacy initialization protocol. Clients may retrieve task-oriented prompts for variant workflows.

A typical starting point is search_variant_evidence, called with the GRCh38 assembly and one public variant query. Supported inputs include exact HGVS expressions, rsIDs, and verified GRCh38 alleles. When a ClinVar or Ensembl record is already available, preserve its transcript version and distinguish submitted ClinVar assertions from the automated Folklore classification.

The Folklore Clinical Variant Interpretation MCP MCP server delegates scientific operations to the hosted service; the companion repository skill does not contain the variant-resolution or evidence-aggregation implementation.

Setup and configuration

No Folklore account or API key is required. Configure an MCP-compatible client with the hosted Streamable HTTP URL, then discover the available tools before making calls. The repository also documents integration routes for Biomni, Biorouter, Dify, n8n, Galaxy, KNIME Analytics Platform, OpenAI Agents SDK, and direct Streamable HTTP clients, but these are client or workflow choices rather than additional requirements for the hosted endpoint.

A companion agent skill is included in the repository for project-scoped, Codex, and OpenClaw installations. It helps route requests involving pathogenicity classification, VUS review, supported variant resolution, ClinVar or population-frequency evidence, variant-linked literature, and gene-disease lookups to the MCP service.

Tools and capabilities

• Query gene-disease associations using a gene symbol or HGNC identifier.
• Search from a disease name or MONDO identifier to associated genes.
• Interpret supported GRCh38 germline variants that are already identified.
• Retrieve available classification criteria, evidence, provenance, and source versions.
• Find literature linked to variants and other supported biological records.
• Use task-first prompts and an optional support_helena helper where exposed by the service.

Limitations and notes

The service is public and read-only. It does not ingest raw DNA, FASTQ, BAM, or VCF files; perform variant calling; analyze a patient genome; or process batches. It accepts no patient, phenotype, family, segregation, or private case context.

Outputs require review by a qualified professional and are not diagnoses, treatment recommendations, or independent evidence of clinical validity. The public benchmark described by the project is a reproducibility and comparison protocol, not independent clinical validation. A VUS review or literature result should therefore be treated as an evidence-gathering step rather than a final clinical determination.