Agent Orchestrator: Supervise Teams of Coding Agents from Planning to Merge

Introduction

Agent Orchestrator (AO) is a local desktop workspace crafted to manage and supervise teams of coding agents from the initial planning stages through to code merge. It addresses the complexities of running multiple agents across a project, providing a unified environment where every coding task gets its own agent, workspace, and feedback loop. AO ensures a clear overview of all activities, pull requests, CI runs, and reviews through a live Kanban board.

Why Use Agent Orchestrator and Its Benefits

AO transforms agent-driven development by providing a structured approach to managing complex projects.

Unified Workspace for Agent-Driven Development: It tackles the challenges of coordinating multiple agents, such as task delegation, context provision, preventing branch collisions, and tracking changes through review and merge.
Focused Worker Sessions: Each worker in AO represents an isolated unit of execution, dedicated to a single task with its own coding agent and workspace. This isolation ensures clarity and independent understanding for parallel work.
Project-Level Orchestration: The project orchestrator acts as a persistent planning and coordination agent, operating at a higher level than individual tasks. It helps explore ideas, brainstorm approaches, reason through tradeoffs, and convert ambiguous outcomes into concrete plans, leveraging project history and live AO state.
Legible Kanban Board: A live Kanban board provides an operational view of the project, deriving card positions from session, pull request, CI, and review statuses. This helps identify what is moving, what is blocked, and where attention is most needed.
Streamlined Workflow: AO offers a comprehensive workflow from idea to merge, allowing users to start at the right level (worker or orchestrator), delegate focused work, build in isolation, supervise live state, and close the feedback loop effectively.
Key Product Highlights:
- Pull Requests and Agent Reviews: Integrates CI, mergeability, reviewer state, and interactive agent reviews directly within the worker session.
- Agent-Controllable Browser: Provides an isolated in-app browser preview for each worker, preventing state conflicts in parallel UI tasks.
- Native Interfaces, One Supervisor: Supports both structured Chat and agents' native terminal UIs, centralizing task context, workspace state, and feedback.

Installation

To get started with Agent Orchestrator, download the latest desktop application for your platform. AO automatically checks for updates.

Platform	Download Link
macOS (Apple silicon)	Download
macOS (Intel)	Download
Windows	Download
Linux (AppImage)	Download
Linux (Debian/Ubuntu)	Download
Linux (Fedora/RHEL)	Download

After installation, open Agent Orchestrator and point it to the repository you wish to manage. The desktop app handles the daemon, eliminating the need for a separate CLI. For agent CLI setup and troubleshooting, refer to the official installation guide.

Examples

Agent Orchestrator facilitates a seamless workflow from conception to completion. Here's how it works:

Start at the right level: Begin by assigning a clear task directly to a worker, or develop a larger outcome with the project orchestrator to shape the overall plan.
Delegate focused work: Initiate workers yourself or allow the orchestrator to create them, providing the necessary context and ownership.
Build in isolation: Each Git-backed worker operates in its own branch and worktree. Standalone agents can use AO-managed branchless directories without requiring a project or repository.
Supervise live state: AO monitors agent activity, pull requests, CI, review feedback, and merge conflicts, reflecting these updates on the Kanban board in real-time.
Close the feedback loop: Directly inspect any worker, make project-level decisions with the orchestrator, and relay actionable failures or review comments back to the agent responsible for the work.

This integrated approach allows agents to leverage their native strengths while AO provides the essential project context, isolated execution, coordination, and an operational overview.