agent-tower-plugin
A Claude plugin that enables multiple AI agents to collaborate and discuss, providing multi-perspective solutions for complex tasks.
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-tower-plugin. It is used for: A Claude plugin that enables multiple AI agents to collaborate and discuss, providing multi-perspective solutions for complex tasks. Full Skill content: https://321skill.com/skills/agent-tower-plugin/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.
The Agent Tower Plugin addresses the potential limitations or biases inherent in a single AI model's perspective. When tackling complex tasks like code reviews, technology selection, or solution design, a single AI's answer may lack comprehensiveness. This plugin orchestrates multiple AI assistants (such as Claude, Codex, Gemini) to simulate collaborative discussion processes—like parliamentary debates, adversarial debates, or production-review cycles—thereby synthesizing more reliable and in-depth conclusions.
Usage is highly intuitive via the Claude Code command-line interface. You can directly invoke commands like /tower:council, /tower:debate, or /tower:deliberate along with your question. The plugin automatically assigns roles, organizes the discussion flow, and ultimately outputs a synthesized answer after multiple rounds of deliberation. The entire process is interactive: you can specify the number of AI participants, customize roles, and view detailed discussion progress.
This plugin is particularly well-suited for technical professionals who require rigorous decision-making or deep analysis, such as full-stack developers, backend developers, AI agent developers, and project managers. It significantly enhances decision quality when you're conducting architecture design reviews, evaluating the pros and cons of different tech stacks, reviewing complex code for security issues, or need a technical plan validated through multiple checks.
Regarding usage recommendations, ensure your system has at least two available AI CLI tools installed (e.g., Claude Code CLI and Codex CLI) to support the multi-agent collaboration foundation. For critical decisions, it's advisable to use the /tower:council mode and increase the number of participating agents (via the --agents parameter), while enabling the --verbose mode to observe the full discussion process, which helps you understand the reasoning behind the final conclusion.
Key Features
Unlike single AI Q&A or simple prompt chaining, it simulates structured collaborative processes like "parliament" or "debate," mandatorily introducing adversarial and diverse perspectives. Its core is the "multi-agent deliberation" mechanism, not merely prompt optimization or workflow chaining.
Limitations
Not suitable for quick, simple answers to everyday queries, as the multi-round deliberation process consumes more time and tokens.
FAQ
How do I use it after installation?
In Claude Code, simply enter a command like `/tower:council "your question"` to initiate a multi-agent parliamentary discussion.
Do I need to install all AI tools?
Not all, but you need at least two installed and configured (e.g., Claude Code CLI and Codex CLI) to run the multi-agent mode.
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-tower-plugin/raw/index.md to read the original Skill definition (Markdown format) for agent-tower-plugin, and install it according to the instructions.
Raw Markdown URL for AI: /skills/agent-tower-plugin/raw/index.md