rajkumarbestha/openapi-mcp-server

Create MCP Server with OpenAPI Spec File

• Easily turn any OpenAPI spec into a set of LLM-accessible tools using the Model Context Protocol (MCP).

• This example is using https://fakerestapi.azurewebsites.net/index.html API

MCP Architecture

• MCP Hosts: Programs like Claude Desktop, IDEs, or AI tools that want to access data through MCP.
• MCP Clients: Protocol clients that maintain 1:1 connections with servers.
• MCP Servers: Lightweight programs that expose specific capabilities through the standardized Model Context Protocol.
• Data Sources: Your computer’s files, databases, and services that MCP servers can securely access.
• Remote Services: External systems available over the internet (e.g., through APIs) that MCP servers can connect to.

Let's MCPise our APIs now!

Overview

This repository provides an example of a Model Context Protocol (MCP) server that dynamically exposes API endpoints as tools using an OpenAPI specification. It leverages the FastMCP framework to generate a server from an OpenAPI spec (such as a Swagger Bookstore API), and provides both a client and an interactive chatbot interface for interacting with the server using LLM-based agents (via LangChain and LangGraph).

Features

• Dynamic API Exposure: Loads an OpenAPI spec (e.g., open-api-spec.json) and exposes its GET endpoints as callable tools via FastMCP.
• LLM Integration: The client and test scripts use OpenAI models (via LangChain) to interact with the server, allowing natural language queries to be mapped to API calls.
• Bookstore Example: The main example is a Bookstore API, but the architecture is generic and can be adapted to any OpenAPI spec.
• Streaming & SSE: The server uses Server-Sent Events (SSE) for real-time communication with clients.

Main Components

• Dynamic API Exposure: Loads an OpenAPI spec (e.g., open-api-spec.json) and exposes its GET endpoints as callable tools via FastMCP.
• LLM Integration: The client and test scripts use OpenAI models (via LangChain) to interact with the server, allowing natural language queries to be mapped to API calls.
• Bookstore Example: The main example is a Bookstore API, but the architecture is generic and can be adapted to any OpenAPI spec.
• Streaming & SSE: The server uses Server-Sent Events (SSE) for real-time communication with clients.

Main Components

• bookstore_mcp_server.py: Starts the FastMCP server from an OpenAPI spec, exposing endpoints as tools.
• bookstore_mcp_client.py: Example client that connects to the MCP server and invokes tools using an LLM agent.
• bookstore_interactive_agent.py: Interactive chatbot for testing the server with natural language queries.
• pyproject.toml: Project metadata and dependencies.

Getting Started

1. Place your OpenAPI spec (e.g., openapi_spec.json) in the project root.
2. Install dependencies as specified in pyproject.toml.
3. Run bookstore_mcp_server.py to start the server:

   uv run bookstore_mcp_server.py
   

4. Use bookstore_mcp_client.py or bookstore_interactive_agent.py to interact with the server:

   uv run bookstore_mcp_client.py
   # or
   uv run bookstore_interactive_agent.py
   

For more details, see the code and comments in each file.