# togomcp

**Category:** 🧬 Biology & Bioinformatics  
**Repository:** https://github.com/dbcls/togomcp  
**GitHub Stars:** 19  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/togomcp

## Description
MCP server for life-science knowledge graphs

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What togomcp MCP server does

The togomcp MCP server gives AI assistants access to biological and biomedical data from RDF databases and selected REST services. RDF Portal resources are queried with SPARQL, while supported external services are accessed through API-specific tools. The result is a single MCP interface for searching, retrieving, and relating information across life-science sources.

Available data spans proteins, genes, chemistry, lipids, pathways, disease, literature, microbiology, glycomics, ontologies, taxonomy, and identifier conversion. Named resources include UniProt, PDB, jPOST, NCBI Gene, Ensembl, HGNC, ChEMBL, PubChem, ChEBI, Reactome, ClinVar, PubMed, NCBI Taxonomy, and TogoID. This places the project in the Biology & Bioinformatics category and makes it relevant to agents that need structured biological knowledge rather than general web search.

## How it works

The server presents MCP tools to an AI client. Depending on the resource, a tool sends a SPARQL query to an RDF Portal endpoint or calls a supported REST API, then returns the service response to the client. The repository provides a hosted deployment at `https://togomcp.rdfportal.org/`, so clients can connect without installing the project locally.

A local deployment uses the `togo-mcp-local` entry point and can communicate over stdio. Docker provides an HTTP deployment, with the supplied Compose configuration defining main and test services. When a deployment sits behind a reverse proxy, the host allow-list and forwarded-protocol settings must be configured consistently with the proxy.

## Setup and configuration

For local installation, use Python 3.11 or newer and the `uv` package manager. Clone the repository, enter its directory, and run `uv sync`. Claude Desktop can launch the local process by invoking `uv` with the project directory and `togo-mcp-local` arguments. The configuration must pass `NCBI_API_KEY` when NCBI tools are needed.

Docker users can build the provided image and run it directly on port 8000, or use `compose.yaml` to start the main and test services. The Compose setup reads image tags, ports, and other values from `.env`. Optional tool-call logging is enabled by setting `TOGOMCP_QUERY_LOG` to a writable file path; logging is disabled when the variable is empty or unset.

## Tools and capabilities

The togomcp MCP server includes tool coverage for:

- SPARQL access to RDF Portal life-science databases
- REST access to NCBI E-utilities, UniProt, ChEMBL, PDB, Reactome, Rhea, MeSH, and other listed services
- Protein, gene, chemistry, pathway, disease, literature, microbiology, ontology, and taxonomy lookups
- Identifier conversion through TogoID
- Optional KEGG searches, entry retrieval, pathway operations, linking, and conversion

KEGG tools include `kegg_find`, `kegg_get_entry`, `kegg_pathway_graph`, `kegg_pathway_neighborhood`, `kegg_pathway_paths`, `kegg_pathway_cycles`, `kegg_link`, and `kegg_conv`.

## Limitations and notes

NCBI API access requires an NCBI API key for the NCBI tool group. KEGG is disabled by default and appears only when both `TOGOMCP_ENABLE_KEGG=1` is set and the local stdio entry point is used. The README states that KEGG access is intended for academic users at academic institutions, and the hosted service does not enable it.

Reverse-proxy deployments need an appropriate `TOGOMCP_ALLOWED_HOSTS` value and trusted forwarded-IP configuration. The proxy must also provide `X-Forwarded-Proto`; otherwise HTTPS requests can be interpreted as plain HTTP. Query logging can include request metadata and, only when explicitly enabled, raw client IP addresses, so its privacy implications should be considered before activation.

The togomcp MCP server does not require local installation when using the hosted endpoint, but local and container deployments require access to the relevant upstream databases and APIs.

_Full upstream README: https://allmcps.com/mcp/togomcp/readme_