I wasted a morning and gained an opinion: Conductor is still the best tool out there if you want to orchestrate multiple coding agents.

Why would you want to move beyond Claude Code or Codex on their own? Running documentation at MotherDuck means going through sometimes tens of PRs a day from other people while also working on new features. A visual overview of which agent is doing what is essential in that workflow, as is a fresh, curated context for each agent and the ability to see which files have been touched.

What I looked at this morning:

• Solo (Aaron Francis): a couple of features I think are amazing, like task management and notes you can pull into your agent context. But I’m missing the ability to easily spin up worktrees within repos, work with multiple agents on one feature, and see which files are affected. I don’t think the terminal is the future either, but who knows, I might be wrong.
• Xirp (Spotify): very similar to Solo and quite alright for managing multiple sessions in a project. Hard to come back to a session and see what actually changed, and hard to switch between agents (Claude, Codex, etc.).
• Kiro Crew (AWS): nice UI, but more oriented towards managing sessions over time than multiple sessions across multiple repositories.
• ChatGPT/Codex: honestly great project and session management, and better browser use than the Claude Code desktop version. But I like to use both Claude and GPT models simultaneously, and Claude is still better with text for documentation.

Conductor has plenty of downsides too: it uses way too much CPU, there are no tasks or notes, and claude -p means no Claude-in-Chrome. But for now I’ll stick with it.