FAQ

Short answers to the questions we hear most. For anything not covered here, the rest of the docs go deeper — start with Policy gates and Settings & config.

01 Source code & security

Does Leroy see my source code?

Yes — it has to read and write code to do the work. But it runs each job in an isolated, per-job sandbox that's torn down when the job finishes. Your code is checked out into that sandbox for the duration of the work and isn't kept on Leroy's own servers. Access to your repos is scoped to the ones you connect.

Is my data isolated from other accounts?

Yes. Every account is strictly isolated — its tickets, conversations, code context, and memory are scoped to that account and never shared with another. Isolation is enforced in depth, not just in the UI.

02 Control

Can I review changes before they ship?

Yes. The ship gate controls this: set it to Always ask and every change waits for your review and approval on the pull request before it merges. You can set this for the whole account or just for specific repos. See Policy gates.

How do I pause or stop it?

Set the gates to Always ask and Leroy stops acting on its own — work waits in your queue until you approve it. To hold a specific piece of work, file or move a ticket to parkedso Leroy leaves it alone until you're ready. The per-task spend cap and risky-path rules also stop work automatically when they trip.

03 Cost & models

What does it cost?

You pay for the work Leroy actually does, billed per task. Lighter tickets cost less than large, multi-repo changes. You can set a spend cap per task so nothing runs away — see Settings & config. Reach out for current pricing.

Which AI models does it use?

Leroy uses best-in-class models, chosen per stage of the work — and we update them as better ones ship. You don't pick or manage models yourself; Leroy routes each job to the right one.