MCPServer

Implements a model context protocol server using Wolfram Language

Installation

Run this command in your terminal to add the MCP server to Claude Code.

Run in terminal:

Command

claude mcp add --transport stdio rhennigan-mcpserver python -m Wolfram.MCPServer \
  --env WOLFRAM_LICENSE_KEY="your-llm-kit-or-license-key-if-required"

How to use

This MCPServer provides Wolfram Language computation capabilities to MCP clients by exposing Wolfram tools as MCP tools. It supports predefined Wolfram-focused tools such as WolframLanguageEvaluator, WolframAlpha, ReadNotebook, WriteNotebook, SymbolDefinition, and more, enabling AI agents and clients to perform symbol lookup, notebook interaction, and Wolfram Language computation within conversations. The server can be extended with additional Wolfram tools or custom tool suites to suit specific workflows, including semantic search over Wolfram documentation and integration with Wolfram Alpha queries.

To use the server, install and run the MCPServer module (via the Wolfram Language module) and connect an MCP client (e.g., Claude Desktop, Cursor, or another MCP client). Once connected, select the appropriate predefined server configuration (e.g., Wolfram, WolframLanguage, or WolframAlpha) or create a custom server definition, and start the server. The client will then be able to send MCP requests that invoke Wolfram Language computation, semantic search, and notebook read/write operations through the exposed tools.

How to install

Prerequisites:

Wolfram Language 14.2 or higher and a valid license or Wolfram Engine access
Access to install Wolfram MCPServer package (Paclet) or equivalent browser-based distribution

Installation steps:

Install the MCPServer Paclet
- In Wolfram Language: PacletInstall["Wolfram/MCPServer"]
Load the MCPServer package
- In Wolfram Language: Needs["WolframMCPServer"]
Install a server configuration (example for Claude Desktop or general Wolfram Language execution)
- Example (via WL): InstallMCPServer["ClaudeDesktop"]
Run the server (example) depending on your chosen configuration
- If using WL-based server, ensure the Python wrapper is available if your deployment uses the Python bridge, or run the WL server directly using the provided module entry point: RunMCPServer["ClaudeDesktop"]
Connect an MCP client and verify the MCP tab/connection status in the client UI.

Additional notes

Tips and common issues:

Ensure your Wolfram Language license or Wolfram Engine is properly configured before starting the MCPServer.
If you switch between server configurations (e.g., WolframLanguage vs. WolframAlpha), remember to restart the MCPServer so the client sees the updated tool set.
For semantic search tools requiring an LLMKit subscription, confirm subscription status and associated API keys if you enable WolframAlphaContext or related tools.
When deploying in containers or headless environments, expose the MCP port and ensure proper network access for MCP clients.
Environment variables (like WOLFRAM_LICENSE_KEY or related keys) may be required depending on your deployment and licensing; replace placeholders with your actual credentials if needed.