Skip to main content
Welcome to Karada! This tutorial walks you through transforming any software API into a production-ready Model Context Protocol (MCP) server that AI agents and coding assistants can use immediately. What you will accomplish:
  1. Create a project in your team workspace
  2. Ingest an API specification (OpenAPI or Postman) with Auto-MCP
  3. Bolt on composable plugins (Sentry tracing, GA4 telemetry, rate limits)
  4. Deploy the server to Karada Managed Hosting or route via the Unified Gateway

​
Prerequisites

Before beginning, make sure you have:
  • A Karada.ai account.
  • A valid API specification (OpenAPI or Postman URL or file). If you need a sample, test with the public Petstore OpenAPI Spec.
  • An MCP client or harness (such as Cursor, Claude Code, or Windsurf).

​
Step 1: Create a Workspace & Project

A project serves as the container for your MCP server deployments, CI/CD sync webhooks, plugins, and usage tracking.
  1. Log into your Dashboard.
  2. Select or create your Team Workspace.
  3. Click Create Project in the upper right.
  4. Select your ingestion source (OpenAPI URL, File Upload, or Postman Collection).
  5. Enter a project name (e.g., Petstore AI Tools).

​
Step 2: Configure Ingestion & CI/CD Sync

Karada’s Go engine parses your API definition into native tool schemas and wires up continuous delivery.
1

Provide Specification URL or Upload

Paste the URL to your openapi.json / swagger.yaml / Postman collection, or upload the file directly.
2

Configure CI/CD Sync Webhook

Optionally configure a webhook to keep your tools in sync with CI/CD on every API deploy and spec drift with zero downtime.
3

Review Detected Tools

Karada compiles routes into discrete MCP tools. Verify the detected tool names, parameter schemas, and descriptions.

​
Step 3: Attach Plugins & Deploy

With tools confirmed:
  1. Attach Composable Plugins: Toggle 1-click plugins for Sentry error tracing, Google Analytics/GA4 telemetry, token-bucket rate limits, or AI firewalls.
  2. Click Deploy Server.
  3. Select your deployment target:
    • Karada Managed Hosting: Provision a high-throughput stateless Streamable HTTP runtime with sub-5ms latency and automated TLS.
    • Unified Gateway: Route through gateway.karada.ai/mcp to use alongside other team and marketplace tools.
    • Download as Standalone: Run and inspect the server locally or inside a private VPC.

​
Next Steps: Connecting Your AI Client

Once your MCP server is running, connect it to your AI client to start calling tools in conversations: 👉 Connect Claude Code, Cursor, & Windsurf in the Auto-MCP Deep Dive
👉 Configure Plugins & Middleware in the Plugins Guide
👉 Multiplex all tools in the Unified Gateway