The problem with BYOC client for Google docs / Figma is that the underlying data isn't simple plaintext, it's too complex, and the protocol requires things like operational transforms, so BYOC wouldn't make sense.
Clients don't have to have exact same power.
For example it would totally make sense to grep over a google doc represented as a plaintext file.
Or at least traverse over the complex structure to extract some data, even if it's read only.