String Db

What String Db MCP server does

The String Db MCP server connects MCP clients to STRING protein-protein interaction data. It supports research workflows that begin with free-text gene or protein names and continue into interaction analysis, network inspection, functional enrichment, or comparisons across organisms.

The String Db MCP server covers about 12,000 organisms through STRING's JSON API. Human data is the default because the species parameter defaults to NCBI taxonomy ID 9606; other organisms can be selected by supplying their taxonomy IDs.

How it works

The server provides five domain tools:

• resolve maps gene or protein names to STRING identifiers. It accepts a list of identifiers and optional species and result-limit values.
• interactions returns interaction partners for identifiers. Optional parameters control species, minimum required score, result count, and network type.
• network produces a network image URL together with tabular network data. Species, score threshold, and network type can be specified.
• enrichment calculates functional enrichment for a gene set, with an optional species value.
• homology returns homology mappings for identifiers in other organisms.

The remote endpoint is hosted by Pipeworx and uses the STRING data source at https://string-db.org/api/json/. Gateway connections also expose shared Pipeworx meta-tools, including discovery and routing tools. Those additional tools are not part of the local standalone process.

Setup and configuration

For a remote MCP connection, configure the gateway URL in the client:

{
  "mcpServers": {
    "string-db": {
      "url": "https://gateway.pipeworx.io/string-db/mcp"
    }
  }
}

The String Db MCP server can also run locally without a Pipeworx account:

{
  "mcpServers": {
    "string-db": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-string-db"]
    }
  }
}

The local package communicates over stdin and stdout and exposes only the five String Db tools. The remote gateway and local process use the same underlying data source and tool set. HTTP clients can call tools directly through the gateway's /v1/tools/ endpoints.

Tools and capabilities

Use resolve before downstream queries when input contains aliases, symbols, or other free-text names. Use interactions to retrieve partner relationships and network when both a visual network reference and tabular results are useful. enrichment is intended for interpreting a supplied gene set, while homology supports mapping identifiers between organisms.

The gateway also supports ask_pipeworx, which accepts a plain-English question and selects a tool and arguments. This routing capability is available through the gateway connection, not through the standalone local process.

Limitations and notes

No API key is required for the documented STRING access. Remote gateway use does not require an account for the first calls, while the local mode avoids the gateway entirely.

The species argument is an NCBI taxonomy ID rather than a common organism name. Network results depend on the selected required score and network type, so those settings should be chosen for the analysis being performed.

A gateway pack endpoint lists the pack's tools plus shared Pipeworx meta-tools, so its tool count is larger than the five domain tools. The full gateway exposes all catalog packs directly; either gateway route can use ask_pipeworx for broader routing. The project is licensed under MIT.