Pharos MCP Server is an MCP server providing access to the Pharos database, a comprehensive NIH knowledge base for studying the drug genome, proteins, diseases, and chemical compounds in drug discovery research. It lets an AI assistant query this data directly through the Pharos GraphQL API.

Use it when you want an AI assistant to look up a protein target by its UniProt ID, find known targets associated with a disease, look up a chemical compound or drug by identifiers such as ChEMBL ID, or explore drug approval status and biological activity data, instead of querying Pharos's GraphQL API directly. It also supports GraphQL introspection, so an assistant can inspect the Pharos schema itself before writing a query. It is scoped to drug-discovery research data in Pharos specifically, not a general biomedical search tool, so it complements rather than replaces broader literature or protein-database search.

The server exposes a single tool, pharos_graphql_query, which executes GraphQL queries against the Pharos API to retrieve information about targets, diseases, and ligands. Through this tool an assistant can query targets (proteins) using identifiers such as UniProt ID, query diseases by name and find associated targets, query ligands (compounds and drugs) by identifiers such as ChEMBL ID, run GraphQL introspection to understand the Pharos API schema, access information about protein functions and their associated diseases, and retrieve data on chemical compounds including drug approval status and biological activity. Example queries include asking about a protein by UniProt ID such as P05067, known targets for a disease such as Alzheimer's, or a specific ligand such as CHEMBL12.

Connect to a running Pharos MCP server over NPX Remote:

npx mcp-remote YOUR_PHAROS_MCP_SERVER_URL_HERE

For Claude Desktop, configure it as:

{
  "mcpServers": {
    "pharos": {
      "command": "npx",
      "args": ["mcp-remote", "YOUR_PHAROS_MCP_SERVER_URL_HERE"]
    }
  }
}

The server is designed for deployment on Cloudflare Workers and connects to the Pharos GraphQL endpoint at https://pharos-api.ncats.io/graphql, with a server URL in the format https://pharos-mcp-server.your-account-name.workers.dev/sse. Because the tool exposes raw GraphQL access, an assistant that first runs introspection can discover the full Pharos schema and construct more precise queries than the documented examples alone would suggest.

Drug discovery researchers who want an AI assistant to query NIH Pharos data on protein targets, diseases, and chemical compounds directly. It is available under the MIT License with academic citation requirements.