With Coding Kitten, FlackBot, and Eliza, there are at least three precedents for this.
There is no official support, e.g. an API or something like that; any bot has go through the same channels as a normal user. This includes Eliza (which, just for the record, was just a fun weekend project of mine, and nothing I did in the capacity as a dev team member).
There's also no promise that we're not going to break it. If we feel that it's necessary to change an AJAX route or JSON format, we will not care whether this breaks anything but the official JavaScript client. We may be nice and give people a heads-up, but we also may not.
Finally, a bot will be held against abuse standards just like any other user. If it "behaves" inappropriately, we're not going to like that. This holds both as far as the actual chatting (content/flooding) goes, as well as considering communication with the server (if you poll the server for new messages every 50ms, that's going to get you banned pretty quickly).
That said, we're all programmers here, and this is the kind of "exercise" we like. So we'll probably not come after you with pitchforks.
But just to make it clear: We can, at any time and with or without any reason, say "We don't want that anymore; turn it off now."