AWS Labs has released CLI Agent Orchestrator (CAO), an open-source multi-agent orchestration framework that coordinates multiple AI coding CLIs—such as Claude Code, Codex CLI and Gemini CLI—using a supervisor-worker pattern, letting a single developer run several agents in parallel.
October 21, 2025 · AWS Labs · Open Source
CLI Agent Orchestrator: Turning Coding CLIs Into a Multi-Agent Team
CAO coordinates multiple AI coding CLI tools — Claude Code, Amazon Q, Gemini and more — with a supervisor-worker pattern. A central agent keeps project context while delegating focused tasks to isolated specialist workers.
10+
supported CLI providers bound into one team
$0
fully open-source, free to self-host
100%
local execution — no external data beyond the CLIs
How the supervisor-worker pattern flows
SUPERVISOR
keeps full project context
→
DECOMPOSE
split into focused tasks
→
WORKERS
isolated tmux sessions
→
COORDINATE
via MCP: handoff · assign
Architecture · security · performance · testing — each handled by a specialist, while the supervisor avoids polluting workers' context. Humans can attach to any session mid-task (HITL).
What developers value
Parallel execution across branches and PRs
Routing of CI log feedback to agents
Dashboard visibility across the team
Hierarchy beats single-agent setups on decomposition & scale
Where it's still early
Coordination is hard in production
Setup overhead to get started
Needs good prompting and profile design
No public benchmarks yet
Execution modes & control planes
Parallel
Sequential
Swarm
Cron-scheduled
Web UI
CLI
MCP
Requirements: Python 3.10+ · tmux 3.3+ · uv. Persistent memory, per-agent tool restrictions, and webhook plugins for Discord, Slack and more.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…