Creating an MCP Server

This guide walks through creating an MCP server using MCPHero’s Chat UI, using a real-world PostgreSQL-backed example.

We’ll build a server that exposes tools for querying customers and orders, while showing how MCPHero’s structured flow guarantees that the server is deployable, secure, and immediately usable by MCP clients.

---

Example Scenario

Assume you have a PostgreSQL database with the following tables:

customers

orders

The goal is to create an MCP server that allows clients to:

• Look up a customer by name
• Fetch all orders for a given customer
• Fetch all orders created in the last hour

---

Step 1: Describe the Server in Chat

You start by describing what you want in natural language.

Example prompt:

I have a PostgreSQL database with customers and orders. I want MCP tools to:

• find a customer by name
• fetch all orders for a given customer
• fetch all orders created in the last hour

At this stage, the AI focuses only on intent discovery, not code generation.

---

Step 2: Tool Discovery & Selection

Based on your description, MCPHero proposes a set of tools with explicit responsibilities.

Proposed tools:

• get_customer_by_name
• get_customers_orders
• get_last_hour_orders

You can:

• Accept all tools
• Rename them
• Remove tools you don’t need
• Add new ones

This confirmation step prevents accidental or unused tools from being generated.

---

Step 3: Authentication Configuration

Next, you configure how the MCP server authenticates incoming requests.

Common options include:

• Static API key via request header
• OAuth2 provider
• No authentication (internal usage)

Example configuration:

• Header name: X-API-Key

The generated server will automatically enforce this authentication for every request.

---

Step 4: Runtime Values & Database Credentials

Because this server connects to PostgreSQL, MCPHero explicitly asks for the required runtime values.

Example values:

• POSTGRES_HOST
• POSTGRES_PORT
• POSTGRES_DB
• POSTGRES_USER
• POSTGRES_PASSWORD

These values are:

• Injected at runtime
• Never hardcoded into the codebase
• Handled differently depending on deployment type (static vs ephemeral)

---

Step 5: Deploy the Server

Before any tools are exposed, the server itself is deployed first.

Once deployment completes, MCPHero provides:

• A server URL
• Authentication instructions (header name + value)
• Environment status and logs

At this point, the MCP is already live

---

Post-deployment steps

Connect to the Server from MCP clients

After deployment, MCPHero shows ready-to-use connection details along with one-click integrations for popular MCP clients.

You’ll see:

• Server URL
• Required authentication headers
• Tool discovery endpoint

One-click client setup

Convenient buttons are provided to instantly connect the server to:

• Claude
• Cursor
• Other MCP-compatible clients

Try MCP functions in browser interactively.

You can test and call your MCP tools from the server page after it is deployed. Depending on tier, source code is visible.