A local setup for multi-repo Cursor agents

inside Linear are hard to use when a team works across several separate code repositories instead of one monorepo. The agents can miss how the repositories depend on each other, especially around .

A local works better because the agent can see the wider context at once. The setup uses a second local copy of the codebase as an automated work environment, with the same remote Git upstream but separate local Git state.

Linear gets a new status called "Ready for AI," and a script checks Linear once per minute for issues with that status. When an issue is ready, the script updates the development branches across the repositories, creates new branches using Linear's "Git Branch Name," links the Linear issue to the , runs Cursor CLI to implement the issue, and opens a pull request.

Key points

  • can struggle when related code is split across several repositories.
  • A local gives the agent more context than a single-repo setup.
  • Linear is used as the task queue through a "Ready for AI" status.
  • A script checks Linear every minute, prepares branches, links the issue, runs Cursor CLI, and creates a pull request.
  • The approach turns a local machine into an automated .
Read original