Approve what your AI agent does, from your lock screen
An agent that stops to ask you something is not slow. It is just waiting for a person who is not there yet.
You start a coding agent on something real: refactor this, delete that table, ship this change. It runs for a while and then it hits a fork it should not decide alone - overwrite the file, or leave it? Drop the column, or keep it? So it stops and waits. If you are not at the keyboard, it just sits there. Ten minutes, an hour, however long it takes you to notice.
That dead time is the actual cost. Not the run itself - the gap between the moment the agent needs a person and the moment a person shows up.
What lands on your phone
The agent asks a question through a tool call the same way it would call any other tool. That call reaches an app running on your Mac, and the Mac sends a push to your phone. On the lock screen you get a notification with the actual question, and if the agent offered a few choices, real buttons for each one - tap one and you are done. For a yes-or-no like "can I delete this file," it is an Approve and a Deny button, nothing else to type.
Nobody has to unlock the phone for this. You read the question, tap the answer, put the phone back down.
The part that matters more: what does NOT get sent
A push notification has to travel through Apple - that is just how iOS delivers pushes, nobody gets to skip it. So the obvious worry is real: is the question about my code sitting on Apple's servers, or worse, on ours?
The push itself carries none of that. It only carries an ID. When it lands, your phone reaches out to your own Mac directly and fetches the real question, the options, and any image, over your own network. The sentence "delete the users table?" never rides the push - only a pointer that says "go ask your Mac what this one says."
This is also why an agent actually waits for you instead of guessing: the tool call it made does not return until you answer. It is not polling, it is not moving on and hoping you catch up later - it is genuinely parked until a real reply comes back, or it times out. And a timeout is never quietly treated as a yes. If nobody answers in time, the agent gets told no answer came, the same as an explicit deny.
The tradeoff: your Mac has to be the one doing the asking
This only works because the question is answered by an app running on your Mac, not by a server somewhere. That is the whole privacy story, and it comes with a real cost: your Mac has to be on and reachable for any of this to happen. Close the lid on a sleeping laptop and there is no agent running, so there is nothing for your phone to fetch and nothing to answer. This solves being away from your desk. It does not solve being away from the machine entirely.
The other honest limit: this is for the yes-or-no and pick-one moments, not for reading a long diff or steering the agent in a new direction from a notification. For that you still want to look at a real screen, which is a different, bigger job than tapping a button on a lock screen.
Which agents this works with
Any agent that can call tools over MCP, which by now is most of the serious ones - Claude Code, Codex, Cursor, and VS Code among others. The agent does not need to know anything about your phone. It calls a tool named something like "ask" or "approve" the same way it would call any other tool, and everything past that point is between your Mac and your phone.
That also means MightyMouse turns your iPhone into a screen, trackpad and microphone for your Mac is not a coding-agent feature bolted onto a phone app - it is the same phone-to-Mac connection MightyMouse already uses for dictation and remote control, reused for a different kind of message.
Stop losing time to an agent waiting on you.
Your iPhone as a trackpad and mic for your Mac. Free to try.
Common questions
Can I approve a coding agent from my phone without unlocking it?
Yes, if the question came with tappable options - a plain approve/deny or a short pick-one shows up as buttons right on the lock screen. Typing a free-text answer still needs the phone unlocked.
Does the question text get sent through a server somewhere?
The push notification itself only carries an ID, never the question. Your phone fetches the real text, options, or image straight from your own Mac once the push arrives - so the content only ever travels between your two devices.
What happens if I do not answer in time?
The agent gets a timeout, and a timeout is never treated as approval - it comes back as a denial, the same as if you had explicitly said no. Nothing risky runs just because nobody was around to answer.
Does my Mac need to be awake for this to work?
Yes. The agent and the app answering on its behalf both run on your Mac, so it has to be on and reachable. This fixes being away from your desk while the agent runs, not being away from a machine that is asleep or off.
If your agent keeps going idle while you are away from the desk: MightyMouse puts your Mac on your phone - approve from the lock screen, and use the same phone as a trackpad and mic when you need to sit back down and actually drive.