Skip to content

Latest commit

 

History

History
128 lines (91 loc) · 3.28 KB

getting_started.md

File metadata and controls

128 lines (91 loc) · 3.28 KB

Getting Started with OpenAPI2MCP

This guide will help you get started with OpenAPI2MCP, a tool that converts OpenAPI specifications to MCP servers with tools.

Prerequisites

  • Python 3.8 or higher
  • An OpenAPI specification file (JSON or YAML)
  • Basic familiarity with APIs

Installation

Install OpenAPI2MCP using pip:

pip install openapi2mcp

Quick Start Guide

Step 1: Prepare an OpenAPI Specification

You'll need an OpenAPI specification file for the API you want to use. You can either:

  • Use an existing OpenAPI spec file (JSON or YAML)
  • Create a simplified spec for your API
  • Find public specs for popular APIs

For this guide, let's create a simple API specification:

# simple_api.yaml
openapi: 3.0.0
info:
  title: Simple API
  version: 1.0.0
servers:
  - url: https://api.example.com/v1
paths:
  /hello:
    get:
      summary: Get a greeting
      operationId: getGreeting
      parameters:
        - name: name
          in: query
          schema:
            type: string
      responses:
        '200':
          description: A greeting message

Save this to a file named simple_api.yaml.

Step 2: Explore the Tools

First, let's see what tools OpenAPI2MCP can extract from this specification:

openapi2mcp convert --spec-file simple_api.yaml --output tools.json

This will create a tools.json file containing the MCP tools. You can examine it to see how the API endpoints were converted to tools.

Step 3: Start the MCP Server

Now, let's start an MCP server with these tools:

openapi2mcp serve --spec-file simple_api.yaml --port 8000

This will start an MCP server on port 8000. You should see output confirming that the server is running.

Step 4: Test the Server

You can test the server using curl or any HTTP client:

# Get server information
curl http://localhost:8000/mcp

# Get available tools
curl http://localhost:8000/mcp/tools

# Execute a tool
curl -X POST http://localhost:8000/mcp/run \
  -H "Content-Type: application/json" \
  -d '{
    "name": "getHello",
    "parameters": {
      "name": "World"
    }
  }'

Step 5: Connect to an LLM

To use your MCP server with an LLM, you'll need:

  1. An LLM or client that supports MCP
  2. Configure the client to connect to your MCP server URL (http://localhost:8000/mcp)
  3. The LLM should discover the available tools and be able to use them in its responses

Authentication (Optional)

If your API requires authentication, you can set up OAuth credentials:

  1. Create a .env file with the following variables:

    API_CLIENT_ID=your_client_id
API_CLIENT_SECRET=your_client_secret
API_TOKEN_URL=https://api.example.com/oauth/token

  2. When you start the MCP server, it will use these credentials to authenticate with the API.

Next Steps

Now that you have a basic MCP server running, you can:

  1. Integrate with more complex APIs: Use more comprehensive OpenAPI specs
  2. Customize authentication: Implement custom authentication logic for specific APIs
  3. Connect multiple clients: Allow multiple LLMs or clients to use your MCP server
  4. Create client applications: Develop applications that use your MCP server to interact with the API

Check out the full documentation and examples for more information.