Skip to main content
This guide walks you through creating an MCP server, writing a tool, testing it in the dashboard, and connecting it to an AI client.

​
Prerequisites

  • An MCPCore account — sign up free, no credit card required
  • That’s it. No local installs needed.

​
Step 1 — Create a server

1

Open the dashboard

Go to app.mcpcore.io and navigate to My Servers.
My Servers list page
2

Click New Server

Click the New Server button in the top-right corner.Fill in:
  • Name — a short, descriptive name (e.g. github-tools)
  • Description — optional, but helpful for teams
  • Security mode — start with Public for now; you can change it later
Create server form
3

Save

Click Create server. You’ll be taken to the server detail page, where you can see the endpoint URL and start adding tools.

​
Step 2 — Write a tool

1

Scroll down to Tools

On the server detail page, scroll down to the Tools section and click Create New Tool.
2

Name your tool

Give your tool a name. Tool names become part of the MCP schema the AI sees — use lowercase snake_case (e.g. fetch_repo_stats).
3

Add a parameter

Click Add parameter and add:
  • Name: repo
  • Type: string
  • Description: GitHub repository in owner/repo format
  • Required: yes
4

Write the code

Paste the following into the code editor:
const res = await sdk.http({
  method: "GET",
  url: `https://api.github.com/repos/${params.repo}`,
});

const data = JSON.parse(res.body());

return {
  stars:    data.stargazers_count,
  forks:    data.forks_count,
  language: data.language,
  topics:   data.topics,
};
Tool code editor
5

Save the tool

Click Save. The tool is now part of your server.

​
Step 3 — Test it

1

Open the Run panel

With the tool selected, click the Run tab (or the play button in the editor toolbar).
2

Enter a parameter value

In the repo field, enter a real repository name — e.g. anthropics/anthropic-sdk-python.
3

Click Run

Click Run tool. The response panel shows the JSON output within a second or two.
Run panel showing tool output

​
Step 4 — Connect to an AI client

Copy the Endpoint URL from the server detail page and add it to your AI client:

Claude Desktop

Add MCPCore to your claude_desktop_config.json.

Cursor

Add MCPCore via Cursor’s MCP settings panel.

Windsurf

Configure MCPCore in Windsurf’s MCP config file.

Any MCP client

Any client that supports the MCP spec works out of the box.

​
What’s next?

Add secrets

Store API keys and reference them as env.MY_KEY — never hardcode credentials.

Change security mode

Lock your server behind API Key or OAuth 2.0 before going to production.

Query a database

Use sdk.db() to connect to PostgreSQL, MySQL, or SQL Server.

Monitor traffic

See every request your server receives and how fast it responded.