Personal Agent
Most software on your phone waits to be told exactly what to do, one tap at a time. A personal agent is the opposite arrangement: you state an outcome — "sort out my travel for the conference" — and it does the deciding, the looking up and the booking on your behalf.
What makes it personal
The agentic machinery here is ordinary — a plan–act–observe loop with tools attached. What distinguishes a personal agent is the context and the authority it carries:
- Long-lived memory of one individual: preferences, contacts, calendars, past decisions, the fact that you dislike early flights.
- Tools bound to your accounts — mail, calendar, messaging, payments — so acting means acting as you.
- Standing permission within limits you set, rather than a fresh instruction for every step.
Why the stakes are different
The blast radius is your life. A coding agent that goes wrong wastes a branch; a personal agent that goes wrong messages the wrong person or spends real money. Designs lean on approval gates for anything irreversible — payments, deletions, outbound messages — and on narrow permission scoping for everything else. The hard problems here are not reasoning problems but trust problems: how much memory about you should persist, who can read it, and what the agent may do when it is uncertain. Assistants built on the OpenAI Agents SDK or the Claude Agent SDK make those the explicit configuration surface rather than an afterthought.
A personal agent is an agent that acts on behalf of one individual, carrying long-lived memory of their preferences, calendars and contacts and holding tools bound to their own accounts. It accepts an outcome rather than a command — arrange this trip, clear this inbox — and decides the intermediate steps itself. Because acting means acting as the user, it depends on tight permission scoping and approval gates for anything irreversible.
Think of the difference between a search engine and a good personal assistant. You ask a search engine for flight prices; you ask an assistant to sort out your trip — and they go away and do it.
A personal agent is software of the second kind. It knows your calendar, your contacts and your habits, and it can use your email and your accounts to get things done. Tell it the outcome and it works out the steps.
Because it is acting as you, it is kept on a short leash: a sensible one checks with you before spending money or sending anything you cannot take back.