Storm

Storm MCP Server enables LLM applications to seamlessly share context with RAG data sources and tools using a standard Model Context Protocol, while integrating Storm's API and tooling for a plug-and-play RAG experience.

How to use

You connect Storm MCP Server to your LLM client to enable context sharing, tools invocation, and file-based data access. Start the server using the provided startup script, then configure your MCP client to point at the Storm server through the local stdio channel. Ensure the Storm API key is available to the server as described in the startup script so the server can authenticate with Storm services. Once connected, you can send non-streaming chat requests, list available tools and data sources, upload documents, and invoke predefined tool handlers to operate on your data.

Typical usage patterns include starting the server, authenticating with your Storm API key, and then using your MCP client to:

• Query the LLM with shared context from your RAG sources
• Inspect and call available tools such as sending non-streaming chats, listing agents or buckets, and uploading documents
• Upload documents for indexing and retrieval within your RAG pipeline
• Interact with Storm API endpoints to manage resources and data sources

How to install

Prerequisites you need on your machine

• Python 3.8+
• bash-compatible shell (sh)
• Git (optional if you fetch files from a repository)

Prepare and run the Storm MCP Server using the startup script provided in your Storm MCP directory.

# Ensure the startup script is executable
chmod +x /Users/sigridjineth/Desktop/work/storm-mcp-server/scripts/run.sh

# Start the MCP server (uses the script path shown in your setup)
sh /Users/sigridjineth/Desktop/work/storm-mcp-server/scripts/run.sh

Configure the MCP client to connect to Storm MCP Server by adding the following MCP server entry to your client configuration. This runs the server as a local stdio process.

{
  "mcpServers": {
    "storm": {
      "command": "sh",
      "args": [
        "/Users/sigridjineth/Desktop/work/storm-mcp-server/scripts/run.sh"
      ]
    }
  }
}

Note: The startup script expects an API key to be available for Storm integration. Provide the API key in the environment where the script runs, for example by exporting STORM_API_KEY before starting the server, or by configuring it inside the script as shown in your setup.

Additional notes

Project components expose a three-tier MCP architecture: the host (LLM application), the client (protocol implementation), and the server (function providers). Storm MCP Server implements the server layer to expose resources and tools to the LLM.

Key data access and tool capabilities include context sharing, a standardized tool system, file management, and API integrations with Storm endpoints. The server defines and handles tools such as chat actions, agent listing, bucket listing, and document uploads.

Available tools

send_nonstream_chat

Send a non-streaming chat to the LLM, enabling synchronous inquiries with context from data sources.

list_agents

List available agents that can be invoked within the MCP workflow.

list_buckets

List storage buckets available for storing or retrieving data used by RAG workflows.

upload_document_by_file

Upload a document by file path to be indexed and used as a data source in RAG.

tool_upload_file

Upload and handle files via the dedicated file server for MCP tool operations.