Agent Orchestrator is an open-source orchestration layer for managing fleets of parallel AI coding agents working on a shared codebase. It reads repository structure, decomposes implementation work into parallelizable tasks, and runs agents in isolated git worktrees so multiple coding lanes can progress without stepping on the same files.
The project focuses on the operational layer around autonomous coding: assigning tasks, supervising runs, handling CI failures, responding to code review comments, and coordinating pull requests. That makes it closer to a multi-agent engineering control plane than a single coding assistant.
Agent Orchestrator is useful for teams experimenting with Claude Code, Codex, Aider, and similar coding agents because it makes parallel execution, review loops, and human escalation explicit. It does not remove the need for engineering review, but it can reduce the manual coordination burden around many small agent tasks.
The canonical public repository is now hosted under AgentWrapper at github.com/AgentWrapper/agent-orchestrator. Teams should verify the current README, release activity, and setup requirements before adopting it in production automation workflows.
