Workflows for running coding agents in parallel
Conductor workflows for running Claude Code, Codex, and Cursor in separate workspaces, with setup scripts, terminals, diffs, checks, and pull requests.
Use these workflows when one terminal session hides too much state: directory, branch, task ownership, and review path.
Each independent task should have its own workspace: a worktree-backed checkout with its own branch, terminal, diff, checks, and pull request path. Start with the workflow for the agent you want to run, then use the docs links to make each workspace runnable without manual repair.
Agent workflows
Start with the workflow for the agent you want to run. Each page explains where terminal sessions break down, how worktrees isolate tasks, and how Conductor keeps each branch reviewable.
Claude Code workflow
Run parallel Claude Codes
Run independent Claude Code tasks in separate workspaces with their own branches, terminals, diffs, checks, and pull requests.
Codex workflow
Run parallel Codexes
Run independent Codex tasks locally without losing track of each task's checkout, branch, terminal, diff, or review path.
Cursor workflow
Run parallel Cursor agents
Run independent Cursor tasks locally without losing track of each task's checkout, branch, terminal, diff, or review path.
Set up the repo once
Parallel agents work best when each new workspace can run the project without manual repair.