Skip to main content
Auto-MCP is Karada’s flagship feature. It bridges the gap between traditional REST APIs and modern AI agents (like Claude, Cursor, or ChatGPT) by automatically generating a Model Context Protocol (MCP) compliant server.

​
The Problem

AI models are incredibly smart, but they are isolated. They cannot interact with your custom company data, SaaS applications, or internal microservices natively. To fix this, developers usually have to spend hours writing custom integration code to wrap their existing REST APIs into formats the AI can understand.

​
The Karada Solution

Karada eliminates this manual work entirely. If your API has an OpenAPI (v3) or Swagger (v2) specification, Karada can instantly generate a fully-typed TypeScript MCP server that acts as a secure bridge between the AI and your API.

​
How the Pipeline Works

  1. Intelligent Parsing: Karada reads your provided openapi.json or swagger.yaml document.
  2. Schema Translation: It translates complex JSON Schema objects, query parameters, path variables, and request bodies into native MCP Input Schemas using the official @modelcontextprotocol/sdk.
  3. Endpoint Mapping: It maps every API route to a distinct MCP tool. The tool description is automatically derived from your OpenAPI summary and description fields, ensuring the LLM knows exactly when and how to use it.
  4. Code Generation: Karada compiles an executable, zero-dependency Node.js/TypeScript application containing the entire proxy logic.

​
Authentication Handling

Most APIs require some form of authentication (API Keys, Bearer Tokens, Basic Auth). Auto-MCP supports passing these credentials dynamically at runtime. When you start the generated server, you simply pass the required tokens as environment variables: 
# Example: Running a generated server that requires a Bearer token
API_KEY="your-secret-key" npm start
Karada automatically inspects the securitySchemes in your OpenAPI spec and generates code that intercepts all outgoing requests from the MCP server to inject the correct headers or query parameters.

​
Connecting to AI Clients

Once your server is generated and running (either locally or deployed on Karada), you need to configure your AI client to use it.

​
Claude Desktop Configuration

To connect your generated Karada server to Claude Desktop, edit your claude_desktop_config.json file: Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json Add the following configuration, replacing the paths with the location of your unzipped Karada build: 
{
  "mcpServers": {
    "my-karada-api": {
      "command": "node",
      "args": ["/absolute/path/to/your/unzipped/karada-server/build/index.js"],
      "env": {
        "API_KEY": "your-secred-token-if-needed"
      }
    }
  }
}
Restart Claude Desktop. You will now see a new hammer icon (🛠️) indicating that Claude has full access to your API directly through Karada’s generated tools!