Creator Docs

Build and publish MCP servers

Everything you need to build, test, and ship an MCP server on the marketplace.

Getting started

What you need to know before building your first MCP server.

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants call external tools. Your server exposes capabilities that any MCP-compatible client can discover and use.

Read the spec

Supported languages

Official SDKs for TypeScript and Python. Community SDKs available for Go, Rust, Java, and C#.

Prerequisites

  • Node.js 18+ or Python 3.10+
  • A GitHub repository, or a hosted remote endpoint
  • Basic understanding of the MCP protocol

Building your server

From scaffold to working server in four steps.

1

Set up the MCP SDK

Scaffold a new project with the official CLI.

TypeScript

npm create @anthropic-ai/mcp-server

Python

pip install mcp

See the official quick start guides for TypeScript and Python.

2

Define your tools

Register tools with names, descriptions, and input schemas. Your server exposes capabilities that AI assistants can call. Each tool should do one thing well and return structured JSON.

3

Test with MCP Inspector

The Inspector opens a browser UI where you can verify all tools appear, test with sample inputs, and check JSON responses. Free, no API credits needed.

npx @modelcontextprotocol/inspector <path-to-server>
4

Test in a real client

Add your server to Claude Desktop, Claude Code, or another MCP-compatible AI application and try it conversationally. Make sure the AI can discover and use your tools naturally. This is the best way to catch usability issues before submitting.

Submitting to the marketplace

What you need and what happens after you hit submit.

What you need

  • GitHub repository (public or private with PAT), or a remote endpoint URL
  • Server name and description
  • Category selection
  • Pricing model (free, one-time, or monthly subscription)
  • Use cases and tags

What happens next

  • Your submission is validated as a real MCP server
  • Local servers get a multi-layer code scan; remote servers get endpoint probing and trust analysis
  • Servers above the safety threshold are approved automatically
  • Others go to manual review
  • Once approved, your server goes live on the marketplace

Speed up submission with LAUNCHGUIDE.md

Add a LAUNCHGUIDE.md to your repo and the submit form will auto-fill your listing details, tags, and setup requirements. Ask your AI code editor to generate one from your codebase, or download the template.

Both one-time purchase and monthly subscription models are supported.

Security scores are indicators to help you make informed decisions, not guarantees. Always review permissions before connecting any MCP server.

Monetization

List for free or charge for your work. You choose.

Free servers

Always free to list. Great for open-source projects and building reputation in the MCP ecosystem.

Paid servers

Set your price. Choose one-time purchase or monthly subscription. We handle Stripe checkout, license keys, and payouts.

After you're listed

Your creator dashboard gives you full visibility into how your server is performing.

Analytics

Track installs, revenue, and reviews over time.

License keys

View issued keys and revoke for abuse.

Reviews

Read user feedback and respond publicly.

Manage your listing

Update descriptions, pricing, and screenshots anytime.

Build your reputation

Earn a letter grade from A+ to F based on uptime, reviews, security, and more. Displayed on every listing.