There is a MCP SEP that outlines multiple ways, that I hope will sooner or later be accepted: https://github.com/modelcontextprotocol/modelcontextprotocol...

While we are waiting on that to become stabilized, we implemented a inspired/co-evolved way to do that in our tool[0], where you mark individual fields in the request/response schema as being file payloads, so that file exchange can be properly orchestrated by the harness and doesn't pollute the context. We just do inline base64 uploads of the required payloads, which in practice we've seen to work quite will until ~100MB files (which is otherwise also the size limit we usually recommend for file processed).

It's annoying that it's not stabilized yet, but for most bigger customers we've seen, they implement 80% of the MCP servers they connect in-house, so doing adjustments to the tool surface, and metadata has been less of a pain for them than we expected.

[0]: https://erato.chat/docs/features/mcp_servers#file-support

I just implemented this with oauth - basically I have a tool that returns a signed URL that the client can PUT the file to. Works in ChatGPT Work mode. My use-case was to have the client generate an image and upload it.

I have had to hack a couple of workarounds in https://github.com/rcarmo/memento to do uploads, and there's a draft going around, but the general practice in enterprise MCPs seems to be to do it "out of band" and have MCP tools to hand-over storage handles/URLs so the MCP server can do the imports itself "safely".