It is text only, sort of. The ActivityPub protocol is basically just JSON, which is text. It's the Mastodon software that chooses to go above and beyond merely implementing that spec and doing things like fetching and hosting images, making a federated timeline, etc.

What we really need is an ActivityPub protocol reference app. An ActivityPub server that is built to the spec and nothing more. All it does is have inboxes, outboxes, and shuttle the JSON messages back and forth. No web UI. No nothing else. Just ActivityPub. Then it can be entirely up to users to provide their own client software of their choosing to engage with the platform.