iflow-mcp-ai-1st-deepview-mcp 0.2.4

A Model Context Protocol server for analyzing large codebases using Gemini 2.0

DeepView MCP

DeepView MCP is a Model Context Protocol server that enables IDEs like Cursor and Windsurf to analyze large codebases using Gemini's extensive context window.

Features

• Load an entire codebase from a single text file (e.g., created with tools like repomix)
• Query the codebase using Gemini's large context window
• Connect to IDEs that support the MCP protocol, like Cursor and Windsurf
• Configurable Gemini model selection via command-line arguments

Prerequisites

• Python 3.13+
• Gemini API key from Google AI Studio

Installation

Installing via Smithery

To install DeepView for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @ai-1st/deepview-mcp --client claude

Using pip

pip install deepview-mcp

Usage

Starting the Server

Note: you don't need to start the server manually. These parameters are configured in your MCP setup in your IDE (see below).

# Basic usage with default settings
deepview-mcp [path/to/codebase.txt]

# Specify a different Gemini model
deepview-mcp [path/to/codebase.txt] --model gemini-2.0-pro

# Change log level
deepview-mcp [path/to/codebase.txt] --log-level DEBUG

The codebase file parameter is optional. If not provided, you'll need to specify it when making queries.

Command-line Options

• --model MODEL: Specify the Gemini model to use (default: gemini-2.0-flash-lite)
• --log-level {DEBUG,INFO,WARNING,ERROR,CRITICAL}: Set the logging level (default: INFO)

Using with an IDE (Cursor/Windsurf/...)

1. Open IDE settings
2. Navigate to the MCP configuration
3. Add a new MCP server with the following configuration:

   {
     "mcpServers": {
       "deepview": {
         "command": "/path/to/deepview-mcp",
         "args": [],
         "env": {
           "GEMINI_API_KEY": "your_gemini_api_key"
         }
       }
     }
   }
   

Setting a codebase file is optional. If you are working with the same codebase, you can set the default codebase file using the following configuration:

{
   "mcpServers": {
     "deepview": {
       "command": "/path/to/deepview-mcp",
       "args": ["/path/to/codebase.txt"],
       "env": {
         "GEMINI_API_KEY": "your_gemini_api_key"
       }
     }
   }
 }

Here's how to specify the Gemini version to use:

{
   "mcpServers": {
     "deepview": {
       "command": "/path/to/deepview-mcp",
       "args": ["--model", "gemini-2.5-pro-exp-03-25"],
       "env": {
         "GEMINI_API_KEY": "your_gemini_api_key"
       }
     }
   }
}

4. Reload MCP servers configuration

Available Tools

The server provides one tool:

1. deepview: Ask a question about the codebase
   • Required parameter: question - The question to ask about the codebase
   • Optional parameter: codebase_file - Path to a codebase file to load before querying

Preparing Your Codebase

DeepView MCP requires a single file containing your entire codebase. You can use repomix to prepare your codebase in an AI-friendly format.

Using repomix

1. Basic Usage: Run repomix in your project directory to create a default output file:

# Make sure you're using Node.js 18.17.0 or higher
npx repomix

This will generate a repomix-output.xml file containing your codebase.

2. Custom Configuration: Create a configuration file to customize which files get packaged and the output format:

npx repomix --init

This creates a repomix.config.json file that you can edit to:

• Include/exclude specific files or directories
• Change the output format (XML, JSON, TXT)
• Set the output filename
• Configure other packaging options

Example repomix Configuration

Here's an example repomix.config.json file:

{
  "include": [
    "**/*.py",
    "**/*.js",
    "**/*.ts",
    "**/*.jsx",
    "**/*.tsx"
  ],
  "exclude": [
    "node_modules/**",
    "venv/**",
    "**/__pycache__/**",
    "**/test/**"
  ],
  "output": {
    "format": "xml",
    "filename": "my-codebase.xml"
  }
}

For more information on repomix, visit the repomix GitHub repository.

License

MIT

Author

Dmitry Degtyarev ([e-mail])