Connecting Linear
In Workspace settings, open the AI page, choose Linear under Integrations, and complete the authorization.
Conversion asks to read issues and comments, post comments, and update issue status. The agent appears in Linear as its own app user, so its comments are clearly attributable rather than posted as whoever connected it.
A Conversion workspace connects to one Linear workspace, and connecting a different one replaces it. You do this once, and then configure individual agents in Conversion.
Which teams the agent can see
Team access is set in Linear, not in Conversion, and it starts at all public teams, so issues in a private team reach nobody until you say otherwise. Nothing happens and no error is raised, which makes it an easy one to lose time on. A Linear workspace admin changes it:- In Linear, go to Settings → Applications.
- Open Conversion Agent.
- Under Team access, choose a scope, name the teams, and press Update.
This is separate from an agent’s trigger filters. Team access decides which issues Linear will tell us about at all; filters decide which of those a given agent picks up.
Choosing which issues an agent picks up
A Linear trigger has an event type and, optionally, filters.Event types
Comment added is usually the best choice. It only fires when a person deliberately comments, and the same trigger handles their follow-up messages later in the conversation, including a simple “go ahead” after the agent has shown you its work.
Issue created fires once, at creation. Use it only if the label and any files are guaranteed to be there the moment the issue is filed.
Filters
Filters narrow which issues the agent cares about. Without them, it responds to every issue in the workspace.
A filter selects specific things in your Linear workspace (this team, these two labels) rather than matching on how they’re spelled. Labels, projects, and workflow states each belong to a team in Linear, so a filter on a team also narrows which of them are worth choosing.
All the filters you set must match; within one filter, any value you picked matches. So label
Bug or Urgent, and team Engineering.
Working with files
Files attached to an issue, or to a comment on it, are available to the agent. It can read spreadsheets and CSVs, transform them, and produce new files for you to download. See Files and sandbox. This is the core of the pattern: someone attaches a partner file to a ticket, tags it, and the agent does the processing that used to be manual.What the agent writes back
Both are locked to the issue that started the session. An agent can’t comment on or move a different ticket, even if asked to, so an agent scoped to one label can never touch issues outside it.
Several agents, one issue
Each agent with a matching trigger gets its own session on the same issue. Three agents can work one ticket independently, each with its own conversation and its own comments. Later comments reach all of them, because a comment always names the issue it’s on. If you’d rather one agent handled a given kind of ticket, scope them with different labels.Things to know
It won’t reply to itself
The agent’s own comments and status changes don’t wake it up.Pausing doesn’t stop an issue it’s already on
An agent already working an issue keeps responding to new comments there, even paused, because a linked issue accepts the comment without checking triggers. To stop that, delete the agent or disconnect Linear. See Triggers.Filters are checked when the agent joins, not on every comment
If an issue’s status or labels change after the agent has picked it up, it carries on regardless.Related
Agents overview
What an agent is and what you configure on it.
Files and sandbox
What an agent can do with an attached file.