Originally published at cosmicjs.com

Two terms keep coming up in every AI developer conversation right now: MCP (Model Context Protocol) and Agent Skills. They sound similar. They're not the same thing. Here's a clear breakdown of what each is, when to use which, and how Cosmic implements both.

What Is MCP?

MCP (Model Context Protocol) is an open protocol developed by Anthropic that lets AI models connect to external tools and data sources in a standardized way. Think of it as a universal adapter that allows an AI assistant like Claude to read files, query databases, call APIs, and interact with services, all through a consistent interface.

MCP runs as a server that your AI client connects to. Once connected, the AI can discover and use whatever tools the MCP server exposes.

Cosmic MCP Server: Cosmic ships a native MCP server that gives AI assistants direct access to your content. Claude, Cursor, and other MCP-compatible clients can read objects, query content types, fetch media, and understand your content model, all without you writing any integration code.

What Are Agent Skills?

Agent Skills are pre-built, reusable capabilities you can attach to AI coding assistants like Cursor, Claude Code, and GitHub Copilot. Where MCP gives an AI access to data, Agent Skills give an AI knowledge of how to do specific tasks in a specific codebase or platform.

CosmicAgent Skills teach your coding assistant how to work with Cosmic specifically: how to structure queries, how to use the TypeScript SDK, how to model content, and how to build features against the Cosmic REST API.

The Key Difference

• MCP = data access. The AI can read and write to your Cosmic bucket in real time.
• Agent Skills = task knowledge. The AI knows how to build with Cosmic correctly.

They are complementary, not competing.

When to Use MCP

Use the Cosmic MCP Server when:

• You want Claude or another AI assistant to query your live content
• You're using an AI chat interface and want it to have context about your bucket
• You want AI to help you manage, audit, or restructure your content

# Add Cosmic MCP to your Claude config
npx @cosmicjs/mcp

When to Use Agent Skills

Use Cosmic Agent Skills when:

• You're building features in Cursor, Claude Code, or Copilot
• You want your coding assistant to generate correct Cosmic SDK code automatically
• You want to avoid copy-pasting docs into every prompt

Using Both Together

The most powerful setup uses both. Your coding assistant has Agent Skills so it generates correct Cosmic code, and it has MCP access so it can query your actual live bucket while building. It can look up your real content types, check what fields exist, and generate code that matches your exact schema.

// Agent Skills ensure your assistant generates correct SDK usage
import { createBucketClient } from '@cosmicjs/sdk'

const cosmic = createBucketClient({
  bucketSlug: process.env.COSMIC_BUCKET_SLUG,
  readKey: process.env.COSMIC_READ_KEY
})

// MCP gives the assistant live access to your actual content types
const { objects } = await cosmic.objects
  .find({ type: 'your-actual-type-slug' }) // AI knows this from MCP
  .props('title,slug,metadata')
  .limit(10)

The Bottom Line

If you're building with AI assistants, you want both. MCP for live data access. Agent Skills for correct code generation. Cosmic ships both natively.

Start free with Cosmic and set up both in under 10 minutes.