Enterprise AI MCP Server

An MCP (Model Context Protocol) server that provides tools for enterprise-scale AI assistance across indexed repositories and documentation. This server enables AI agents to efficiently search and retrieve code context for development tasks.

Features

• Search: Natural language queries across indexed repos/docs, tailored to a provided technology stack

Installation

1. Clone the repository:

   git clone https://github.com/bitovi/enterprise-ai-mcp.git
   cd enterprise-ai-mcp

2. Install dependencies:

   npm install

3. Build the server:

   npm run build

4. Run the server:

   npm run start

Configuration

Environment

This server requires a webhook endpoint for search. Create a .env file (see .env.example) and set:

WEBHOOK_URL=https://your-webhook-endpoint

MCP Client Settings

To use this MCP server with an MCP client (e.g., VS Code + Cline), add a server entry in your client’s MCP settings file (path varies by client and OS):

{
  "mcpServers": {
    "enterpriseCode": {
      "command": "node",
      "args": ["/path/to/enterprise-ai-mcp/build/index.js"]
    }
  }
}

The server will be loaded automatically by the MCP client.

Prompting Files

The repository includes four prompting files to guide AI interactions and development. These files all include a statement that the agent must read back to you so that you may confirm that the agent has used the file.

• AGENTS.md
• .github/copilot-instructions.md
• .github/prompts/sampleSearch.prompt.md
• .github/instructions/sample.instructions.md

Tools

search

Purpose: Given a natural-language query and technology stack, return the best matches across all indexed repos/docs. Results are provided by the configured webhook.

Inputs:

• Message (string, required): A detailed natural-language description of the user’s query, task, or intent.
• Stack (string, required): Comma-separated list of relevant technologies (e.g., TypeScript, React, Node.js).

Returns: JSON payload returned by the webhook (passed through as text).

Development

• Source code: src/index.ts
• Build output: build/index.js
• The server forwards tool requests to WEBHOOK_URL and returns the JSON response.