I am very excited to share that Conductor is joining Meta!
We have some big, crazy plans for the next interface. If you liked Conductor, you’re gonna love what’s coming.
We’re super excited to work with @natfriedman and @alexandr_wang at MSL to build it. They’ve been on an absolute tear, and I’m convinced MSL has the highest talent density of all the labs.
A huge thank you to our earliest believers @aaron_epstein, @ilyasu, and @nabeel. And to everyone here who tried Conductor, sent us feedback, and prompt requests - you made this so much fun. Thank you ❤️🚂
Choo choo!
When orchestrating it is breaking work up into phases and tasks. It then coordinates the completion of each task using separate sessions.
Those sessions are now hyper focused on completing that task. It then assigns a reviewer session, any findings are then returned to orginal implementator for fixing. You can define how many review loops and your overall flow by just tell the coordinating session.
This makes it capable of completing large features in technically one run by coordinating 100s of sessions/agents in series based on defined flow and gates.
So to answer your question it is very effective.
Bring the agent you already use.
Claude Code and Codex connect through native adapters. Gemini and 39 more agents connect through ACP.
No model of our own. No markup on tokens.
One prompt in. A plan out.
Claude Code splits the feature: subagents on search, Codex on the empty state in its own worktree, tests on the Linux build server. The Plan tab shows every task live.
Meet Glueprint, the Agentic Development Environment.