home / mcp / grok mcp server

Grok MCP Server

Provides chat, completions, embeddings, and model management endpoints for Grok AI via a fast MCP server.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "brewmytech-grok-mcp": {
      "url": "http://localhost:8080/stream",
      "headers": {
        "GROK_API_KEY": "YOUR_GROK_API_KEY"
      }
    }
  }
}

Grok MCP Server enables chat, completions, embeddings, and model operations with Grok AI through a fast, modular MCP setup. It exposes a streaming HTTP endpoint by default and supports multiple operation types with clear error handling and real-time responses.

How to use

You connect your MCP client to Grok MCP Server to perform chat interactions, text or image-based chat inputs, embeddings, and model management tasks. Use the provided tools to list models, inspect a model, generate chat completions or text completions, and compute embeddings. You can enable streaming to receive partial results as they are generated and you can mix text and images in chat sessions for richer interactions.

How to install

Prerequisites you need before installing:
- Node.js installed (recommended LTS version)
- npm (comes with Node.js) or another package manager you use
- A Grok API key if you plan to access Grok services over the cloud
- Optional: a local environment for development and testing

Additional setup and notes

Configuration enables you to connect to remote MCP endpoints and to pass your Grok API key for authenticated access. By default, the server exposes an HTTP streaming endpoint on port 8080. You can run the server locally and point MCP clients to http://localhost:8080/stream when you are ready to test.

Available tools

list_models

List available Grok models and their details.

get_model

Retrieve information for a specific model by its ID.

create_chat_completion

Generate a chat-based completion with optional streaming and sampling controls.

create_completion

Generate a text completion with optional streaming and sampling controls.

create_embeddings

Create vector embeddings from input text using a specified model.