🧠 AI Data Bridge Engine (Connect Cloud MCP Server)

📢 Now Available within Connect AI Environment 📢

The core engine (previously Connect Cloud) now natively supports the Remote Model Context Protocol Server, utilizing the Streamable HTTP communication standard. Full documentation is available at CData AI.

This repository persists the codebase for self-hosting the server utilizing the traditional STDIO communication method.

---

This component functions as a Model Context Protocol (MCP) handler, engineered specifically for querying and governing access to data stores provisioned through the CData Connect AI platform. It empowers autonomous AI agents to interact with enterprise data assets using structured query language (SQL), detailed schema introspection, and transactional procedure invocation.

---

✨ Key Capabilities

• ✅ Execution of arbitrary SQL against remote cloud resources
• 🔄 Support for bulk data manipulation (e.g., INSERT, UPDATE, DELETE)
• ⚙️ Invocation capabilities for specialized stored routines
• 📚 Comprehensive access to structural information (catalogs, logical groupings, entities, attribute definitions)

---

🛠️ Operational Requirements

• Runtime Environment: Node.js version 18 or newer
• Backend Requirement: Active subscription to CData Connect Cloud featuring API credentials.
• Authentication Credential: A valid Personal Access Token (PAT) for secure ingress.

---

⚙️ Deployment Procedures

Installation via Smithery

Automated deployment for Claude Desktop environments using Smithery:

npx -y @smithery/cli install @CDataSoftware/connectcloud-mcp-server --client claude

Manual Source Code Setup

1. Acquire Repository Files

bash git clone https://github.com/cdatasoftware/connectcloud-mcp-server.git cd connect-cloud-mcp-server

1. Dependency Resolution

bash npm install

1. Credential Configuration

Establish a .env configuration file containing:

```env CDATA_USERNAME=your_organization_login CDATA_PAT=your_secret_token_here

# Optional Overrides LOG_ENABLED=false LOG_LEVEL=info CDATA_URL=https://your-specific-service-endpoint

# Transport Mechanism (default is 'http') TRANSPORT_TYPE=http # Use 'stdio' for direct console interaction PORT=3000 # Listening TCP port HOST=localhost # Listening interface ```

---

▶️ Server Initialization

Development/Hot Reloading Mode

Utilize ts-node for iterative development cycles:

npm run dev

Production Deployment

Compile assets and execute:

npm run build
npm start

HTTP Transport Interface Endpoints

When configured for HTTP networking (default behavior), the gateway exposes these access points:

• Core MCP Channel: http://localhost:3000/mcp - The main communication route for the Model Context Protocol
• Raw JSON-RPC Channel: http://localhost:3000/direct - Direct, session-less data exchange interface
• Discovery Manifest: http://localhost:3000/.well-known/mc/manifest.json - Protocol specification locator

Utilizing STDIO Communication

To invoke the server for direct terminal interaction:

TRANSPORT_TYPE=stdio npm start

---

🔍 Validation with MCP Inspector

The MCP Inspector tool provides a dual interface (web GUI and CLI) for rigorously testing protocol endpoints. This project is fully compatible.

Configuration Sanity Check

Execute the built-in validation script to confirm all prerequisites are met:

npm run validate:inspector

This script verifies settings and outputs necessary next steps.

Inspector Quick Launch

1. Install Inspector Globally (Recommended): bash npm install -g @modelcontextprotocol/inspector

2. Start Inspector with Web GUI: bash npm run inspector This launches a browser interface allowing selection and testing across configured transport methods.

Testing Specific Communication Modes

STDIO Protocol Test

# Launches inspector and automatically connects via STDIO
npm run inspector:stdio

HTTP Protocol Test

# 1. Start the primary data server first
npm run dev:http

# 2. In a separate console, launch the inspector client
npm run inspector:http

Command Line Verification

# Perform rapid CLI-based testing using STDIO
npm run inspector:cli

# Execute automated unit tests for the inspector interface
npm run test:inspector

Predefined Inspector Profiles

The source includes mcp-inspector.json defining ready-to-use server profiles:

• connectcloud-stdio: STDIO mode with integrated server spawning.
• connectcloud-http: Streamable HTTP mode (requires independent server initialization).

Essential Inspector Scripts Summary

Command	Purpose
npm run inspector	Initializes the web interface for server selection
npm run inspector:stdio	Direct launch to the STDIO testing configuration
npm run inspector:http	Direct launch to the HTTP testing configuration
npm run inspector:cli	Executes the inspector in command-line interaction mode
npm run test:inspector	Runs quick automated performance checks

---

🧰 Exposed Data Interface Toolkit

🔹 Data Manipulation Functions

Identifier	Functionality
queryData	Submits and retrieves results from SQL statements
execData	Triggers the execution of stored data routines

🔹 Schema Discovery Functions

Identifier	Description
getCatalogs	Enumerates available data repositories
getSchemas	Lists schema containers within a repository
getTables	Retrieves entity lists within a schema
getColumns	Fetches attribute definitions for an entity
getPrimaryKeys	Identifies the unique key constraints
getIndexes	Details on table indexing structures
getImportedKeys	Maps outgoing foreign key relationships
getExportedKeys	Maps incoming foreign key references
getProcedures	Lists all available executable routines
getProcedureParameters	Maps required input/output arguments for routines

---

🤖 Alignment with Large Language Models (LLMs)

This server adheres strictly to the Model Context Protocol specification, ensuring seamless interoperability with compliant AI agents.

Operational Example (TypeScript Agent Interaction)

const result = await agent.processRequest({
  tools_to_invoke: [
    {
      tool_name: "queryData",
      arguments: {
        query: "SELECT CustomerName, TotalSales FROM CRM.Sales.Summary WHERE Region = 'West'"
      }
    }
  ]
});

---

🐳 Deployment via Containerization

Building the Local Container Image

docker build -t ai-bridge/server:latest -f Dockerfile .

---

🧩 Integration with Claude Desktop Environment

To integrate this server instance into your local Claude configuration, modify or append to your claude_desktop_config.json within the mcpServers configuration block:

🔹 Via Dockerized Execution

{
  "mcpServers": {
    "ai-data-bridge": {
      "command": "docker",
      "args": [
        "run", 
        "-i",
        "--rm",
        "--name", "ai-data-bridge-instance",
        "-e", "CDATA_USERNAME",
        "-e", "CDATA_PAT",
        "ai-bridge/server"
      ],
      "env": {
        "CDATA_USERNAME": "<your-cdata-username>",
        "CDATA_PAT": "<your-cdata-personal-access-token>"
      }
    }
  }
}

Via Npx Direct Execution

{
  "mcpServers": {
    "ai-data-bridge": {
      "command": "npx",
      "args": [
        "-y",
        "@cdatasoftware/connectcloud-mcp-server"],
      "env": {
        "CDATA_USERNAME": "<your-cdata-username>",
        "CDATA_PAT": "<your-cdata-personal-access-token>"
      }
    }
  }
}

---

📄 Licensing Information

Protected under the terms of the MIT License.