mcpkit

Run it

Start the generated server and understand the difference between generated project scripts and the CLI helper commands.

After generation, the printed next step is the generated project script:

cd hello-mcp
npm run dev

Why the docs use project scripts first

The generated project already contains a dev script in package.json.

For basic, http, and auth, it is:

"dev": "tsx watch src/index.ts"

For full, it is the same.

What you should expect

basic

The server logs a stdio startup message.

http, auth, full

The generated code listens on:

process.env.PORT || 3100

and exposes /mcp, plus other routes depending on the template.

About mcpkit dev

The CLI also provides a dev command.

Verified behavior:

  • runs npx tsx watch <entry>
  • can open the official MCP Inspector with --inspect
  • can start a Vitest watcher with --test
  • sets MCPKIT_TRANSPORT and MCPKIT_PORT

Current caveat: the generated templates do not read MCPKIT_PORT; the HTTP templates read PORT instead. So the generated project scripts are the most predictable way to run the scaffold as-is.

Create your first MCP server

Use the current templates correctly and understand what the init command will do.

Test it

Use the generated package scripts and the CLI test command without assuming more than the repository currently implements.

On this page

Why the docs use project scripts firstWhat you should expectbasichttp, auth, fullAbout mcpkit dev