Skip to main content
Karada enables LLM-powered agents and coding assistants to safely interact with your proprietary backend systems. This playbook walks through configuring an AI agent to use a Karada-deployed MCP server to read data and perform authenticated actions on your behalf.

​
Prerequisites

  • An existing REST API with an OpenAPI v3 or Swagger v2 specification.
  • An active Karada.ai account.
  • An AI client supporting the Model Context Protocol (e.g., Claude Desktop, Cursor IDE).

​
Phase 1: Ingesting the API Specification

To allow an LLM to understand and invoke your system, expose your API routes as discrete tools:
  1. Log into your Karada Dashboard and create a new Auto-MCP project.
  2. Provide your API specification:
    • Paste a public URL (e.g., https://api.yourcompany.com/openapi.json).
    • Or upload your .json / .yaml file directly.
  3. Karada parses each endpoint into a typed MCP tool, with descriptions derived from your summary annotations.

​
Phase 2: Deploying the Cloud Endpoint

Your generated server runs continuously as an isolated microservice:
  1. Click Deploy from your project overview.
  2. Karada compiles the Go engine runtime, binds plugins, and provisions an isolated container.
  3. Upon build completion, copy your secure HTTPS / SSE URL (e.g., https://my-api.karada.ai/sse).
[!NOTE] Ensure runtime authentication credentials (such as API_KEY or BEARER_TOKEN) are configured in Project Settings → Environment Variables. Karada injects these into outbound requests at the proxy layer.

​
Phase 3: Client Configuration

Connect your AI assistant to your Karada endpoint.

​
Option A: Claude Desktop Setup

Open your Claude configuration file:
  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
Add your Karada endpoint using the Streamable HTTP/SSE transport client:
Restart Claude Desktop. Verify that the Tools (🛠️) icon in the chat bar lists your endpoints.

​
Option B: Cursor IDE Setup

  1. Open Cursor Settings (Cmd + , on macOS, Ctrl + , on Windows).
  2. Navigate to Features > MCP Servers.
  3. Click Add New MCP Server.
  4. Set:
    • Name: company-api
    • Type: command
    • Command: npx -y @modelcontextprotocol/client-cli connect https://my-api.karada.ai/sse
  5. Click Add. Cursor Composer will now automatically call these tools during code generation and debugging.

​
Phase 4: Executing Agent Workflows

With tools connected, prompt your AI assistant directly with operational requests:
  • “Look up user ‘alice@example.com’ and report their current billing status.”
  • “Fetch open issues tagged with ‘urgent’ and draft a bug triaging summary.”
The model reasons over the tool definitions, generates typed arguments matching your OpenAPI schema, executes the call through Karada, and synthesizes the response in real time.