Run agents in parallel

ComputerWeb app3 min readUpdated October 3, 2026

Several conversations can work at the same time, each with its own agent. Give them separate tasks, keep overlapping changes in worktrees, and check in when they need you.

Each conversation runs its own agent, so you can have one building a page while another researches a question. Claude Code’s guide suggests running sessions in parallel “to speed up development, run isolated experiments, or start complex workflows.” The catch: agents in different conversations don’t know about each other.

What to run side by side

  • Separate tasks. A new page in one conversation, an unrelated bug fix in another.
  • Work that doesn’t change files. Research, plans, and reviews can run next to anything.
  • A builder and a reviewer. Claude Code’s guide suggests one session to write and a fresh one to review, since a new agent “won’t be biased toward code it just wrote.”
  • Different agents. Each conversation has its own agent settings, so you can give one task to Claude Code and another to Codex.

How to do it in Enjoy

  1. Start each task in its own conversation. Select New conversation for each. See Use one conversation per task.
  2. Separate overlapping work with worktrees. Enjoy doesn’t stop two conversations from editing the same files. If two tasks touch the same pages, ask one agent to work “in a new worktree,” a separate copy of the project. Worktrees need your project to use Git. To start another conversation in an existing worktree, open the branch button in the message box of a new conversation and choose it under Worktrees.
  3. Add follow-ups to the right conversation. If you think of more for a task that’s already running, send it there. While the agent works, pressing Enter adds it to that conversation’s queue. Don’t start a second conversation on the same task.
  4. Let an agent split the work. You can ask: “Start separate conversations to write the About page and the FAQ.” An agent can start up to five conversations per request, with the same or narrower permissions, and replies with a link to each.
  5. Keep track. The Working filter in the conversation list shows what’s running, and Needs you shows what’s waiting on you. See Stay in the loop without watching.

An example

On a Monday morning you start three conversations:

  • Claude Code: “Build the pricing page from the ‘Pricing plan’ doc.”
  • Codex: “In a new worktree, try a simpler navigation menu.”
  • Any agent: “Compare three newsletter services for a small studio. Don’t change any files.”

While they work, you answer one agent’s question about plan names and review the research when it’s done.

Common pitfalls

  • Two agents on the same files in the same folder. They can overwrite each other’s work. Codex’s guide warns against running tasks on the same files without worktrees.
  • Switching branches while agents work. Enjoy blocks it until the work in the main project folder finishes, and shows Work in progress.
  • Running more than you can review. Every result still needs checking. See Check the agent’s work.
  • Forgetting the cost. Each running conversation uses your AI plan. See Make your agent usage last.
Did this help?

Still need a hand?

Email help@enjoy.dev with your computer, your Enjoy version, and what happened. Leave out passwords and sign-in codes. For questions and ideas, join our Slack community .