Linear Agent

Linear Agent reads your real workspace data, including teams, sub-teams, initiatives, projects, milestones, cycles, issues, comments, and activity history. That means its answers reflect what's actually in Linear rather than generic project advice. Ask about a specific project and it can trace how the work is organized before responding. Real data in, real answers out.

You don't fill out a form to make an issue. Describe what you need in a chat and the agent creates it, then keeps adjusting it as you talk. The same goes for updates to projects, milestones, and initiatives. It turns a multi-field ticket into a sentence. That's the whole trick.

Mention @Linear in any comment and the agent joins that thread with the surrounding context already in hand, so it can edit or reply to its own comments as a discussion evolves. This is the fastest path when the work is already underway and you just want a hand inline. No new chat required.

The dedicated agent chat shows its progress as it works instead of going silent and returning a finished answer. You can watch it move from a question to a draft to concrete next steps, which matters when you need to trust what it changed. No black box.

Linear supports several agent chats at once, each shown as a tab in the toolbar with a short label. Each tab flags unread updates or shows when the agent is still working, so you can run one chat on a launch plan and another on a support backlog without losing your place. Multitasking, handled.

Linear Agent operates inside your existing permissions. It can only reference or change content you already have access to, which is the difference between a helpful teammate and a security headache. Nothing it does is outside what you could do yourself.

A workspace admin or owner can toggle Linear Agent off entirely from Settings, then AI, then Linear Agent. That gives cautious teams a clean way to pilot it or turn it back on later without hunting through configs. Admins stay in control.