awslabs/cli-agent-orchestrator

Multi-agent orchestration for AI coding CLIs — Claude Code, Kiro, Codex, and more, coordinated in isolated tmux sessions

What it solves

It solves the problem of coordinating multiple AI coding CLIs, allowing a supervisor agent to delegate tasks to various specialist agents that can work either in parallel or in sequence.

How it works

CAO runs a local server (cao-server) that launches provider CLIs in isolated terminal sessions. These agents remain full CLI processes, retaining their native authentication and capabilities, while the supervisor uses a set of tools to manage and coordinate the workers.

Who it’s for

Developers who use multiple AI coding tools (such as Claude Code, GitHub Copilot CLI, or Cursor CLI) and want to orchestrate them into a unified workflow.

Highlights

  • Multi-Provider Support: Compatible with a wide range of CLIs including Kiro, Claude Code, Codex, Antigravity, Hermes, Kimi, GitHub Copilot, OpenCode, Cursor, and Grok Build.
  • Flexible Control Planes: Can be managed via a Web UI, shell CLI, or an MCP server.
  • Advanced Workflow Management: Supports scheduled runs, multi-step pipelines, and the ability to turn workflow outcomes into lessons via a self-learning loop.
  • Persistent Memory: Features cross-session memory and reusable "skills" for agent guidance.

Related

  • Project
    getpaseo/paseo

    A unified, self-hosted interface for managing multiple AI coding agents like Claude Code and Copilot across desktop, mobile, and CLI.

  • Project
    mco-org/mco

    A CLI-first orchestration layer for AI coding agents that allows users to run tasks across multiple models in parallel to compare perspectives and reduce blind spots.

  • Project
    multica-ai/multica

    An open-source workspace that orchestrates multiple AI coding agents, allowing users to assign them tasks like teammates and track their progress on a unified board.

  • Project
    Untrivial-ai/agent-orchestrator

    A local desktop workspace for planning, running, and supervising multiple coding agents in parallel, providing isolated workspaces and a live Kanban view of project progress.

  • Project
    aannoo/hcom

    A CLI tool that enables different AI coding agents to message, observe, and spawn each other across terminals to coordinate complex development workflows.