Agent Orchestrator (AO) is a local desktop workspace for running several coding agents on one repository at once. Each unit of work is a worker: one task, one coding agent and model, and one isolated workspace. Git-backed workers get their own branch and worktree, while scratch workers get AO-managed branchless directories, so parallel work does not collide or collapse into a shared context.
Above the workers sits a persistent project orchestrator — a planning and coordination agent scoped to the repository. You use it to explore an idea, weigh approaches, and turn an ambiguous outcome into a concrete plan; it then breaks the plan into tasks, spawns or redirects workers, passes each the context it needs, and follows their progress. The orchestrator owns planning and delegation; workers own implementation, tests, commits, and pull requests.
Every worker appears on one live Kanban board whose columns are derived from session, pull request, CI, and review facts: working, needs you, in review, and ready to merge. Each card carries the task, agent, branch, activity, and pull request state, and opens onto the conversation or native terminal, changed files, the pull request summary and reviews, and an isolated browser preview. AO watches the pull requests it opens and routes failed checks and review comments back to the session that owns the branch.
A local daemon behind the desktop app tracks agent activity and source-control state; the desktop app starts it, so no CLI is required. AO supports 26 coding agents with per-project choice, and you can work with each through structured chat or its own terminal UI. A companion mobile app pairs with the desktop over LAN or Tailscale to watch sessions and open a terminal while execution and code stay on your machine. Builds are available for macOS, Windows, and Linux.
Features
- Isolated workers: every Git-backed task gets its own branch and worktree; scratch workers get branchless AO-managed directories
- Project orchestrator: a persistent planning agent that keeps goals, decisions, and constraints, then delegates tasks to workers
- Live Kanban: working, needs you, in review, and ready to merge, derived from session, pull request, CI, and review state
- Feedback loop: failed checks and requested changes route back to the agent that owns the branch
- Agent choice: 26 coding agents supported, selected per project
- Dual interfaces: structured chat or the agent's own terminal UI, with task context kept in one place
- Agent-controllable browser: per-worker isolated browser profiles for previewing and inspecting a worker's local app
- Pull request view: CI, mergeability, reviewer state, and interactive agent reviews beside the worker
- Mobile companion: pair over LAN or Tailscale to watch sessions and get notified, with execution staying local
- Local daemon: the desktop app runs it, so no CLI setup is required
- Cross-platform builds: macOS on Apple silicon and Intel, Windows, and Linux as AppImage, deb, or rpm
Integrates with
Aider
Command-line LLM pair programmer that maps your repo, edits files, and auto-commits each change to git
Claude Code
Terminal-based agentic coding tool that reads your codebase and runs tasks through natural-language commands
Cline
Open-source coding agent with one engine behind VS Code, JetBrains, a headless CLI, an SDK, and a parallel Kanban board
Codex
Lightweight coding agent that runs in your terminal