How to use prompts and resources in MCP servers? Can prompts and resources be reused or called from tools?

[pedrojfds]

I'm building an MCP server using @modelcontextprotocol/sdk with TypeScript and integrating it into an app like VSCode. From what I see, only tools are recognized and used by the app/client, while resources like prompt are not.

I'm trying to understand:

How are prompt resources meant to be used in an MCP server?

Can a prompt be invoked from within a tool? (e.g., a tool that internally calls a prompt and returns its result)

What's the recommended pattern for reusing prompts across tools or other resources?

In environments like VSCode, is there any support for exposing non-tool resources (like prompts or resources) or linking prompts and tools in some way?

Some documentation or example code would be really helpful to clarify how prompts and other resources should be structured and composed with tools.

Thanks a lot!

[mogulsstar]

same question