Create Your Alter-Ego
A version of you, trained on your work. Private or public, yours to shape.
An alter-ego is not an "AI agent" or an "AI employee." It is a version of you, trained on the writing you have already done, the decks you have already shipped, the calls you have already taken, and the emails you have already sent. It does the work you would do, in the voice you would use, with the judgment you have built.
XO gives you two ways to spin one up: from scratch on your own data, or from a public template you fork and then feed with your data.
From scratch
Train a private alter-ego on your own writing, decks, calls, and emails.
From a template
Fork a public alter-ego, then feed it your data to make it yours.
Publishing
Share an alter-ego publicly so other people can fork it.
Private vs public
There are two flavors of alter-ego, and the difference is who owns the data underneath.
Private alter-egos are trained on data only you own. "You for Memos" learns from every investment memo you have written. "You for Deals" learns from your deal notes, term sheets, and back-channel emails. "You for Inbox" learns from the way you actually reply, not the way LinkedIn thinks you reply. They live in your workspace. Nobody else can use them, fork them, or see what is in them.
Public alter-egos are community-shared templates anyone can fork. "The Patient Investor" carries the disposition and questions of a seasoned LP. "The Mentor" carries the tone of a generous operator who has already seen this movie. "The Strategist" carries the move-three-steps-ahead lens. None of them are trained on your data. They are a starting point: a personality, a framework, a vocabulary. The moment you fork one and start feeding it your work, it stops being a template and becomes a private alter-ego of yours.
Forking a public alter-ego does not pull your data into anyone else's workspace, and it does not push the original author's data into yours. You get the shape; you fill it.
When to spin a new one
The instinct is to make one alter-ego that does everything. Resist it. The alter-egos that earn their keep are narrow.
- You have a recurring piece of work that takes the same shape every time (a weekly memo, a monthly board update, a deal one-pager). Make an alter-ego for that surface.
- You have a body of writing or recordings that captures a distinct mode of yours (negotiating, mentoring, writing for the team versus writing for LPs). Make an alter-ego for that mode.
- You are trying out a new disposition you do not yet have (longer time horizon, more patient, more direct). Fork a public alter-ego in that direction and let it lean against your defaults.
The test is simple. If two alter-egos would draw from substantially different source material and produce substantially different output, they should be two alter-egos. If not, one is enough.
How they run, under the hood
This part is optional reading. If you only want to use alter-egos, skip it.
Under the hood, every alter-ego is a configured agent on the XO Coworker workspace, brokered by xo-cowork-api and run on a runtime like Claude Code, Codex, or OpenClaw. The "alter-ego" label sits on top of the same agent infrastructure documented in the XO Cowork API section. If you want to look under the hood, build a custom alter-ego with structured tools, or wire one into your own frontend, the API docs are the place to start.
Where to go next
Most people start by forking a public alter-ego to get a feel for it, then build a private one against the work they actually want help on.
- New here? Fork a template.
- Have your own material to train on? Start from scratch.
- Built one worth sharing? Publish it.