使用说明
GWAS Catalog MCP
A Model Context Protocol (MCP) server that lets AI agents and agentic IDEs query the GWAS Catalog in plain language — searching traits, studies, and variant–trait associations from published genome-wide association studies.
You don't need to install anything. A hosted server is live at:
https://www.ebi.ac.uk/gwas/mcp
The server is also listed in the BioContext registry.
Connect
The server uses MCP's Streamable HTTP transport. Connect to the hosted endpoint with an MCP 2.x-compatible client.
Codex
codex mcp add gwas-mcp --url https://www.ebi.ac.uk/gwas/mcp
Use codex mcp list to check the configured servers.
Claude Code
For MCP 2.x, use Claude Code's http transport for the hosted Streamable HTTP
endpoint:
claude mcp add --transport http gwas-mcp https://www.ebi.ac.uk/gwas/mcp
Add --scope user to make it available in all your projects.
ChatGPT desktop app
Open Settings → MCP Servers → Add server, choose Streamable HTTP, and enter:
https://www.ebi.ac.uk/gwas/mcp
Save the server and restart the app.
Other MCP clients that support remote Streamable HTTP servers can connect using
the same URL. Clients that only support local STDIO servers can use the
mcp-remote bridge.
What you can ask
The server gives your agent three tools:
| Tool | What it finds |
|---|---|
| Traits | EFO traits (diseases, phenotypes) by name, gene, or publication |
| Studies | GWAS studies by trait, ancestry, gene, or accession |
| Associations | Variant–trait associations with statistics (p-value, odds ratio, …) |
It also exposes reference data (ancestry labels, cohort IDs, variant consequences) that agents use to interpret results.
You don't call these directly — just ask your agent questions like:
- "What EFO trait covers type 2 diabetes?"
- "Find GWAS studies of type 2 diabetes in East Asian cohorts."
- "What variants are associated with LDL cholesterol near the APOE gene?"
- "Show the strongest associations for study accession GCST000001."
Run it yourself
Most people should use the server hosted above.
The best way to test and develop the server locally is to run MCP Inspector in STDIO mode.
The development endpoint is https://wwwdev.ebi.ac.uk/gwas/mcp. Both development
Helm paths are /gwas/mcp; an older deployed path containing /api requires a
Helm release update using the current values.
License
Apache-2.0
