For one, because it breaks the Unix philosophy of "doing one thing and doing that well".
In that vein, I do want Firefox to develop/allow/improve an interface so that machines, amongst which AI-MCPs, can drive my firefox. And do so safely, secure, contained, etc.
So that my AI agent can e.g. open a Firefox tab and do things there on my behalf. Without me being afraid it nukes all my bookmarks, and with me having confidence in safety nets so that some other tool or agent cannot just take over my gmail tab and start spamming under my account.
Point is: I really think Mozilla and Firefox have a role to play in the AI landscape that's shaping up. But yet another client to interact with chatbots is not that. Leave that to people building clients please: do one thing and do it well.