agent-teams
Orchestrating Claude Agent Teams via File System Protocol
Install & Use
Copy this prompt and send it to your AI assistant (Claude / Cursor / TRAE / Codex / WorkBuddy etc.) to auto-install:
Help me install this AI Skill: agent-teams. It is used for: Orchestrating Claude Agent Teams via File System Protocol Full Skill content: https://321skill.com/skills/agent-teams-x-3/raw/index.md Read that page and install it.
The prompt includes a link to the full Skill content. You can also view the full content.
This Skill addresses the lack of a lightweight coordination mechanism when multiple Claude processes need to collaborate in parallel. In real-world development, developers often need to run multiple Claude instances to handle different modules simultaneously. However, manually managing communication and task distribution is cumbersome and prone to conflicts or oversights.
Usage is straightforward: simply create a team configuration file, then launch each Claude process in a tmux pane with a specified identity. Subsequently, each Agent exchanges messages via agreed-upon JSON files, with fcntl locks ensuring concurrency safety—no database or network layer required.
It is ideal for teams or individuals engaged in parallel coding, multi-agent code review, large-scale refactoring, or naturally parallelizable tasks. This is especially true for developers already using the Claude CLI who wish to enhance collaboration efficiency without introducing additional infrastructure.
We recommend using this tool when tasks have inherent parallelism, such as synchronously modifying multiple independent modules or simultaneously conducting code review and fixes. Note that it is primarily suited for Unix/Linux environments (dependent on tmux), and is not recommended for single-file or small tasks, as the coordination overhead may outweigh the benefits.
Key Features
Unlike frameworks such as CrewAI or AutoGen, which rely on a network layer or database, agent-teams operates entirely on a file system protocol. It requires no daemon processes or external services to be started, resulting in a lighter deployment with no additional dependencies. However, this approach necessitates manual management of each Agent process lifecycle.
Limitations
Requires tmux and the Claude CLI, needs the environment variable CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 to be set, does not support Windows systems, and file I/O may become a performance bottleneck during high-frequency interactions.
FAQ
What dependencies need to be installed?
You need to install tmux (for managing processes in panes) and the Claude CLI, and ensure `claude --version` is available. Additionally, you must set the environment variable `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`.
How do I launch team members?
Launch each Agent process using the command: `claude --agent-id <id> --agent-name <name> --team-name <team> --agent-color <color> --parent-session-id <session> --agent-type <type> --model <model>`.
Installation guide for AI assistants
If your AI coding assistant (Claude Code, Cursor, TRAE etc.) can see this page, send it this message to auto-install:
Visit https://321skill.com/skills/agent-teams-x-3/raw/index.md to read the original Skill definition (Markdown format) for agent-teams, and install it according to the instructions.
Raw Markdown URL for AI: /skills/agent-teams-x-3/raw/index.md