openapi-mcp-server

MCP.Pizza Chef: janwilmake

The openapi-mcp-server is a Model Context Protocol (MCP) server designed to enable AI models like Claude and Cursor to interact with and explore OpenAPI specifications using simple, natural language. It simplifies navigating complex API definitions by identifying the relevant OpenAPI identifier, summarizing it in accessible language, and pinpointing the necessary endpoints for the task. This server integrates with oapis.org and supports streamlined API exploration, making it easier for developers and AI agents to understand and utilize APIs without deep technical knowledge of OpenAPI specs.

Use This MCP server To

Search OpenAPI specs using natural language queries Summarize complex API definitions in simple language Identify relevant API endpoints for specific tasks Enable AI agents to explore APIs dynamically Integrate OpenAPI exploration into AI workflows

README

OpenAPI MCP Server

smithery badge janwilmake/openapi-mcp-server context

A Model Context Protocol (MCP) server for Claude/Cursor that enables searching and exploring OpenAPI specifications through oapis.org.

The MCP works by applying a 3 step process :

  1. It figures out the openapi identifier you need
  2. It requests a summary of that in simple language
  3. It determines which endpoints you need, and checks out how exactly they work (again, in simple language)

Important

OpenAPI MCP has found a new owner and has been migrated from v1.2 to v2, which works different to the previous version. You can still access any version prior to v2.0.0 and their README is here

OpenAPI MCP v2 is a Work In Progress and focuses on exploration and providing context about APIs. It does not allow executing the endpoints as tools directly, as authentication isn't a solved problem with MCP yet. However, it's great for codegen!

Expect bugs. Open To Contributers, DM

Features

  • Get an overview of any OpenAPI specification
  • Retrieve details about specific API operations
  • Support for both JSON and YAML formats
  • Tested with Claude Desktop and Cursor

Installation

Installing via Smithery

To install openapi-mcp-server for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @janwilmake/openapi-mcp-server --client claude

Installing via npx

Run and follow instructions:

npx openapi-mcp-server@latest init

Usage in Claude

Once installed, you can ask Claude to:

  • "Find information about the Stripe API"
  • "Explain how to use the GitHub API's repository endpoints"

Claude will use the MCP server to:

  1. First get an overview of the requested API
  2. Then retrieve specific operation details as needed

Requirements

  • Node.js >= 16.17.0
  • Claude Desktop, Cursor, or any other MCP client.

License

MIT

openapi-mcp-server FAQ

How does openapi-mcp-server simplify OpenAPI exploration?
It translates complex OpenAPI specs into simple language summaries and identifies relevant endpoints for easier AI interaction.
Can openapi-mcp-server work with multiple AI models?
Yes, it is designed to work with models like Claude, Cursor, and can be adapted for others such as OpenAI's GPT-4 and Anthropic's Claude.
What is the three-step process used by openapi-mcp-server?
It identifies the OpenAPI identifier needed, summarizes it in simple language, and determines which endpoints to use.
Is there a demo available to see openapi-mcp-server in action?
Yes, a live demo is available at https://x.com/janwilmake/status/1903497808134496583.
How does openapi-mcp-server integrate with oapis.org?
It uses oapis.org as a source to fetch and explore OpenAPI specifications dynamically.
What kind of APIs can openapi-mcp-server handle?
It can handle any API described with OpenAPI specifications, regardless of complexity.
Does openapi-mcp-server require manual configuration for each API?
No, it automatically figures out the OpenAPI identifier and relevant endpoints based on the query.
Where can I find the source code for openapi-mcp-server?
The source code is available on GitHub at https://github.com/janwilmake/openapi-mcp-server.