Skip to main content

Create an MCP Server

The MCP Studio wizard walks you through four steps to create and deploy your server.

Step 1: Add sources​

Paste URLs for the content you want your MCP to reference. Supported source types:

  • Documentation sites — e.g., https://docs.docker.com
  • GitHub repos — public or private (requires GitHub sign-in for private repos)
  • Any website — blog posts, wikis, knowledge bases
Large docs sites

If a docs site has more than 100 pages, the wizard will let you know and recommend either:

  • Providing a more specific URL (e.g., https://docs.docker.com/engine/)
  • Connecting the docs GitHub repo directly for full coverage

Step 2: Choose tools​

Select between 3 and 10 tools for your MCP server. Each tool gives AI agents a specific capability. See the full Tools Reference for details on each one.

Step 3: Review​

Preview your server name, sources, and selected tools. Make any final adjustments.

Step 4: Deploy​

Hit Deploy and your MCP server is created in seconds. You'll receive:

  • A Server URL for JSON-RPC over HTTP
  • An SSE Endpoint for MCP clients using Server-Sent Events
  • A JSON config snippet ready to paste into your AI tool

Your server starts indexing sources in the background. It's usable immediately — it uses live fetching as a fallback until indexing completes.