How to install and use Claude Code Agent Teams (Reverse-engineered)

2026-02-07 · Source: AI Jason · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, medium

Summary

Cloud Code has introduced a significant "agent teams" feature in version 2.1.34, allowing three to five Cloud Code instances to collaboratively complete tasks and projects. Unlike previous sub-agent functionality, these new agent sessions can communicate with each other using various protocols. The main agent orchestrates the process by calling a "team create" tool, which generates a configuration file and an empty team member array, and then a "task create" tool to log a list of to-do items with dependencies. Subsequently, the main agent uses an upgraded "task tool" to spin up new Cloud Code sessions, each inheriting team information and a task list. These sub-agents utilize a "task update" tool to modify task statuses and a "send message" tool for one-on-one or broadcast communication. The communication protocol involves injecting new user messages into each agent's conversation history via an "inbox" folder, and the main agent can send "shutdown requests" to terminate sub-agent sessions and ultimately use a "team delete" tool to close the entire team.

Key takeaway

For AI Engineers and Machine Learning Engineers developing complex problem-solving agents, Cloud Code's new agent teams feature offers a robust framework for distributed task execution. You should explore this capability to tackle intricate debugging scenarios or projects requiring multiple perspectives, as it allows agents to collaboratively investigate, debate findings, and converge on more comprehensive solutions than single-agent approaches. Consider integrating a terminal multiplexer like tmux or iTerm2 to monitor agent activities effectively.

Key insights

Cloud Code's new agent teams enable collaborative, multi-agent task execution with inter-agent communication and shared context.

Principles

• Orchestrate multi-agent workflows via a lead agent.
• Enable agents to autonomously update tasks.
• Facilitate direct agent-to-agent communication.

Method

The main agent creates a team and task list, then spins up multiple Cloud Code sessions. Sub-agents update tasks and communicate via message tools, with the main agent managing shutdown.

In practice

• Update Cloud Code to v2.1.34.
• Set "cloudcode.experimental.agentTeams" to 1.
• Use `cloud-teammate --mode` for split-view.

Topics

• Cloud Code Agent Teams
• Multi-Agent Systems
• Agent Communication
• Task Orchestration
• AI Debugging

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Claude Code Just Got WAY More Powerful — Anthropic's new features enhance Claude's automation, self-correction, multi-agent collaboration, and memory capabilities.
• The Rise of Cloud Coding Agents — Cloud-based AI coding agents enable asynchronous, autonomous development workflows, enhancing scalability over traditional desktop agents.
• Claude Code Agent View IS INSANE! Huge New Update Introduces /goal, sessions, & More! — Cloud Code's new features centralize agent management and enable autonomous, goal-driven execution for developers.
• Claude Opus 4.6 and GPT-5.3-Codex 30 Mins Apart — AI agent teams are rapidly advancing, demonstrating autonomous complex task execution and self-improvement capabilities.
• 🤖 AI Agents Weekly: Claude Code Review, AutoHarness, Perplexity Personal Computer, Cloudflare /crawl, Context7 CLI, and More — Multi-agent systems and automated constraint synthesis enhance AI agent performance and reliability.
• You Have Never Seen An AI Agent Do This Before (Claude Code) — Autonomous AI agents can perform complex, multi-stage projects from ideation to social media promotion.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Jason.