Claude Code Projects vs. running your own manager agent in the terminal. Which way are you going?
retained text
I've been building an app with Claude Code for a few months, and all of my multi-agent setup lives in the terminal. I use herdr, a terminal multiplexer for coding agents. One Claude session acts as a manager: it picks the next GitHub issue, opens a new pane, starts a fresh claude there, and hands it the issue. The child does the work (branch, tests, review, PR, merge). The manager watches the pane, answers or escalates the child's questions, closes the pane, and moves on to the next issue. Some issues need my local machine: SSH to a Windows test PC, the iOS simulator, screenshots. Projects in Claude Code can now run threads on your own machine through Remote Control, and the coordinator does much of what my manager pane does. As far as I can tell from the docs, Projects works in the desktop app, on the web and on mobile, but not in the terminal CLI. So I'm wondering: Is multi-agent work moving into the GUI, with the coordinator there and the terminal mostly hosting sessions? If you've used Projects with local threads, how did it go? In particular: do you have to click "Allow once" for every local thread, and can you keep it to one thread at a time when they share a repo? If you still run your own manager agent in tmux, herdr or something similar, what keeps you there? For me it's control. My manager follows a fixed script, and the coordinator makes its own calls. I'm planning to try Projects when I start my next version, so I'd like to hear what worked and what didn't before I move things over.
document fetched 2026-11-16 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1xwsaw9/claude_code_projects_vs_running_your_own_manager