Linear’s cross-functional support-to-code path (April 2026)

Source: Rhea Purohit, “How we use Linear Agent at Linear,” April 10, 2026, first-person composite of named internal operators; some tools then under internal testing. This is an operating description, not a single named multi-PR feature’s changelog or impact analysis.

Steps and responsibility. CX converts an Intercom customer email into an issue using Linear Agent, which gathers replies and attachments. Triage intelligence routes feature requests to PM, flags duplicates and links observability context. PM Sid Bhargava asks the agent to cluster overlapping customer requests, reviews its output and reorganizes work; he can supplement PRD/spec and draft engineering issues. Engineer Mingjie Jiang asks about architectural history and owners, delegates implementation but retains the human assignment, may finish locally, and an agent reviews every PR before final approval by a human engineer. Merge marks an issue Done; linked requests reopen in Intercom so CX follows up. In a separate smaller example, PM Sid spots an external-source filter missing from issue lists, asks agent to create issue and code a first pass; engineer Paco Coursey reviews within an hour and change goes live. In another path, a support complaint in Slack becomes an issue, engineer Matthijs Wolting scopes it into small targeted steps, resolves it, and the Slack thread notifies CX to contact the customer.

Limits / open verification. This report does not show plan revision across linked PRs, acceptance-test details, post-merge fixes, customer adoption, tool spend or time spent reviewing. “Done” follows PR merge, not measured customer benefit. The Sid filter fix is concrete but deliberately small, and the account’s large-feature steps are aggregated from practices rather than one verifiable trace. See the engineer’s review workflow, integrating trace question, and role shifts.