Most developers want Claude to act—inspect logs, query production data, interact with live systems—not just discuss them. While MCP enables this, manual setup and configuration create operational friction—Server Installation, Node.js dependency chains, and connection debugging that interrupts flow state. We've seen this setup tax stall DevOps teams and solo technical freelancers—burning billable hours on connection debugging instead of delivering infrastructure value.
This guide skips the manual troubleshooting. We will walk you through creating and connecting an MCP server to Claude Desktop using MCP Express. By the time you finish this 5-minute read, you will deploy a fully functional AI agent that queries databases, audits logs, and responds to infrastructure questions in natural language.
What Is MCP—and Why Connect it to Claude?
Model Context Protocol (MCP) is an open-source standard that enables Claude to connect to databases, APIs, and infrastructure through a unified interface. Instead of building point-to-point integrations for every tool in your stack—AWS CloudWatch, PostgreSQL, Slack, SSH servers—MCP provides a single protocol that scales across services.
For freelancers managing client infrastructure, this transforms Claude from a conversational assistant into an operational agent. You can query live databases, audit cloud logs, execute diagnostic commands, and automate workflows—all through natural language. This means answering client questions and troubleshooting issues without interrupting deep work or burning billable hours on manual queries.
Real-World Example:
Instead of SSH-ing into a client server, pulling logs, and parsing errors manually, you ask:
"Show me all failed deployments in the last 24 hours."
Claude queries the infrastructure via MCP and returns a structured answer—no context-switching, no command recall.
For the full technical specification, see Anthropic's official MCP documentation.
The Fastest Way to Create a MCP Server: MCP Express
We built MCP Express for freelancers and DevOps operators who need MCP's operational power without managing infrastructure complexity. No manual server installations, no dependency debugging—just production-ready servers in minutes.
What You’ll Need
Before you begin, ensure you have
Estimated time: 5 minutes from signup to first query.
Deploy your MCP Server in 5 Minutes
Step 1: Sign Up and Log In
Access the MCP Express Dashboard. This serves as your command center for all AI-to-infrastructure connections.
Step 2: Create a New MCP Server
In your dashboard, navigate to MCP Servers.
Then click on New MCP Server.
Pro Tip for Freelancers: Name your server based on the client or project (e.g., Client-A-PostgreSQL) to keep environments isolated and secure.
Then, click on Create MCP Server.
To manage what Claude can see and do, click View Details on your newly created MCP server. This is your command center for security and functionality.
From this view, you can manage:
- Tool Authorization: Select which specific tools (PostgreSQL, AWS, etc.) are active for this server.
- Secure Credential Storage: Safely store connection strings and API keys. Note: Claude never sees your raw credentials; it only interacts with the authorized tools via the MCP bridge.
- Real-Time Analytics: Monitor activity logs to see exactly what queries Claude is running and when.
Choose what Claude can access:
- AWS CloudWatch
- PostgreSQL, MySQL, DynamoDB, MSSQL
- REST APIs
- SSH
- Slack
- Confluence
- Custom internal tools
You control the permissions and the scope—not the AI. You decide exactly which data is exposed and which actions are permitted.
Connecting Your MCP Server to Claude Desktop?
With your server configured, the final step is to create the bridge that allows Claude to "see" your tools.
Step 1: Provision your Claude Client
In the MCP Express dashboard, navigate to Clients in the sidebar.
Click on Add New Client at the top right corner.
Give your client a name, choose Claude from supported clients, and select your MCP server to connect your client with.
Now, we will link the MCP Express bridge to your local Claude application.
From your dashboard’s Clients list, click the three dots (⋮) in the Actions column and select "Setup Instructions."
From the instructions pop-up, download the MCP Express CLI and add the following snippet to your claude_desktop_config.json.
{
"mcpServers": {
"Your Client Name": {
"command": "node",
"args": [
"<path_to_mcp-express-cli.js>"
],
"env": {
"CLIENT_ID": " ",
"CLIENT_SECRET": " "
}
}
}
}
Where to find your config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
If you’re having trouble finding the claude_desktop_config.json file, you can simply open Claude Desktop and go to Files → Settings → Developer. Here, click on the Edit Config button, and you can access the file.
Step 3: Verify and Launch
For Claude to recognize your new configuration, you must fully exit and restart the Claude Desktop application.
Once restarted, navigate to the settings to ensure your bridge is active: go to Files → Settings → Developer. You’ll see your client name for your MCP server.
Start a new chat. You’ll notice a small "plug" icon or a notification that Claude has access to your tools. Instead of asking for a summary or a joke, give it a real command:
"Check the database for the last five failed transactions and tell me the error codes."
The Result: Claude no longer "hallucinates" or asks for a CSV upload. It reaches through the MCP Express bridge, queries your live system, and returns fact-based answers in seconds.
You're now Operational
You've deployed a production-ready MCP server in under 5 minutes. Claude can now query databases, audit logs, and troubleshoot infrastructure—all through natural language.
Scale this across client projects:
Connect your next database or cloud environment in seconds.
Create Your Next MCP Server
Further Resources:
- Documentation - See every supported integration, configuration option, and code example in one place.
- Contact Us - Got a question before signing up, or just want to talk through your setup? Drop us an email.
- Open a Support Ticket - Already inside the app and something's not working? Open a ticket directly from your dashboard and we'll get back to you as soon as possible.
