MCP Tools and Swagger seems very …. well , very similar
| Topic | MCP Tool | Swagger (OpenAPI) |
|---|---|---|
| Purpose | Let LLMs safely call backend functions. | Document and test REST APIs. |
| Audience | AI models + AI app developers. | Human developers. |
| Describes | Tools (actions), not endpoints. | HTTP endpoints, request/response formats. |
| Used by | ChatGPT, LLM apps, agents. | FrontEnd + BackEnd developers. |
| Transport | HTTP / STDIO / SSE | HTTP/HTTPS. |
| Focus | AI tool integration. | API documentation/communication. |
So what’s the point of having 2 separate interfaces, one for calling from Human using AI and other from Human using programming languages ?
More, what if I want to replicate the from one to another ? Should I copy the MCP code to swagger ? Or the swagger code to MCP ?
More than this, could this be performed automatically ?