Swagger

🚀 swagger-mcp Documentation

swagger-mcp is a tool designed to capture Swagger UI. It extracts the swagger.json file and dynamically generates a standardized MCP tool for MCP clients to select and use at runtime.

📽️ Demo Video

Watch the demo video to understand the project's features:

🙌 Support

If you find this project valuable, please support me on LinkedIn:

• 👍 Like and share my demo post.
• 💬 Leave your thoughts and feedback in the comments.
• 🔗 Follow me for future updates.

Your support will help me reach more people and continuously improve the project!

🚀 Quick Start

Prerequisites

Before using swagger-mcp, ensure the following dependencies are installed:

1. LLM Model API Key / Local LLM: You need access to OpenAI, Claude, or Ollama models.
2. Any MCP Client: (For example, use mark3labs - mcphost).

📦 Installation

Install and run swagger-mcp by following these steps:

go install github.com/danishjsheikh/swagger-mcp@latest
swagger-mcp

Running Configuration

Run swagger-mcp directly using:

swagger-mcp --specUrl=https://your_swagger_api_docs.json

Main parameters:

• --specUrl: Swagger/OpenAPI JSON URL (Required).
• --sseMode: Run in SSE mode (Default: false. If true, run as an SSE server; otherwise, use standard input and output).
• --sseAddr: SSE server listening address in the format of IP:Port or :Port (If empty, it will be automatically completed based on --sseUrl).
• --sseUrl: SSE server base URL (If empty, it defaults to http://IP:Port or http://localhost:Port).
• If both --sseAddr and --sseUrl are set, refer to the specific documentation.
• For other parameters, refer to the tool's instructions.

MCP Configuration

Configure mcphost in the .mcp.json file as follows:

{
  "swagger-mcp": {
    "url": "https://your_swagger_api_docs.json",
    "sseMode": false,
    "sseAddr": ":8080"
  }
}

Demo Process

The commands to run the backend, Ollama, and MCP client are as follows:

# Backend service
go run main.go

# Ollama service
ollama serve --model gpt-3

# MCP client
go run client.go

Flowchart

Refer to the link for the project flowchart: Flowchart Link

Areas for Improvement

Users have mentioned the following areas to be improved:

• Tool Definition: A more standardized way of tool definition is needed to improve usability.
• LLM Behavior Control: Enhance the ability to control LLM models to ensure the output meets expectations.
• Context Management: Optimize the context processing mechanism to improve conversation coherence.
• Error Handling: Strengthen the exception detection and recovery mechanism to improve the tool's robustness.

These improvements will make swagger-mcp more perfect. Your feedback and suggestions are welcome!