apo1397/mcp-server-influencer-search

MCP Server Demo

This project demonstrates an MCP (Multi-Cloud Platform) Server designed to find influencers based on various criteria using the Modash API and Gemini API for natural language query processing.

What is an MCP Server?

An MCP Server is a specialized server that exposes a set of tools (functions) that can be called by an AI agent or other clients. It acts as an extension of the agent's capabilities, allowing it to perform complex tasks by leveraging external APIs and custom logic. In this project, our MCP Server provides a tool to search for influencers.

Project Setup

To set up and run this MCP Server, follow these steps:

1. Prerequisites

• Python 3.11 or higher
• uv (a fast Python package installer and runner)
• API Keys for Modash and Gemini (Google Generative AI)

2. Installation

1. Clone the repository:

   git clone <repository_url>
   cd mcp-server-influencer-search
   

2. Install dependencies using uv:

   uv pip install main.py
   

   This will install mcp[cli], google-generativeai, google-api-core, and requests as defined in the main.py file.

3. Environment Variables:

Create a `.env` file in the root directory of the project and add your API keys:

```dotenv
MODASH_TOKEN="your_modash_api_key"
GEMINI_API_KEY="your_gemini_api_key"
```

Replace `your_modash_api_key` and `your_gemini_api_key` with your actual API keys.

4. Running the MCP Server

Manual Run (for testing)

You can run the MCP server manually from your terminal:

cd <project_directory>
/opt/homebrew/bin/uv run main.py

This will start the MCP server, making its tools available.

MCP Server Configuration for AI IDEs (Trae, Cursor, Claude Code)

To integrate this MCP Server with AI IDEs like Trae, Cursor, or Claude Code, you need to configure them to recognize and connect to your server. This typically involves creating a .claude.json (or similar) file in your project's root directory.

Here's an example of how to configure it for Claude Code (the structure might be similar for Trae and Cursor):

.claude.json example:

{
    "mcpServers": {
      "Influencer-Search": {
        "command": "/opt/homebrew/bin/uv",
        "args": [
          "run",
          "<path_to_your_project>/main.py"
        ],
        "env": {
          "MODASH_TOKEN": "${MODASH_TOKEN}",
          "GEMINI_API_KEY": "${GEMINI_API_KEY}"
        }
      }
    }
  }

Explanation of the configuration:

• name: A unique name for your MCP server (e.g., "Influencer-Search"). This is how the AI agent will refer to your server.
• command: The executable command to run your server. In this case, it's the uv executable.
• args: A list of arguments passed to the command. This specifies how uv should run your main.py as an MCP server.
• cwd: The current working directory where the command will be executed. This is crucial for the .env file to be picked up correctly.
• env: An object containing environment variables that will be set when the server starts. It's recommended to reference variables from your .env file using ${VAR_NAME} syntax for security and maintainability.

Important Notes for IDE Integration:

• Restart IDE: After creating or modifying the .claude.json file (or equivalent configuration), you will likely need to restart your AI IDE (Trae, Cursor, Claude Code) for the changes to take effect.
• Environment Variable Loading: Ensure your IDE is configured to load environment variables from your .env file. The cwd setting in the MCP server configuration helps with this.
• Tool Discovery: Once the server is running and configured, the AI IDE should automatically discover the tools exposed by your main.py (e.g., get_influencers_from_modash, get_location_id) and make them available for use.