agent-tower/core

README preview

**English** | [简体中文](./README.zh-CN.md)

# Agent Tower

> The command center for your army of agents.

## Highlights

### 1. Kanban, Isolated Workspaces, and Multi-Agent Support

Agent Tower supports a range of agent backends, including Codex, Claude Code, Gemini CLI, Kimi CLI, Cursor CLI, and more, so you can use whichever one fits the job. Its task status system (**Review**, **Running**, **Done**, and more) makes every task's state easy to understand, while an automatically isolated workspace keeps each task from interfering with the others.

<p align="center">
  <img src="./docs/assets/agent-tower-demo.gif" alt="Choosing an agent, running a task with Codex, and watching the task move from Running to Review" width="100%" />
</p>

### 2. Team Mode (Real Multi-Agent Collaboration)

In Team mode, agents work together in a shared group chat. Each agent remains an independent member, so one agent's long context cannot contaminate the others or send the task off course. You can designate a lead to break down the task, an implementer to make the changes, a reviewer to inspect quality, and a tester to verify the result. When the work is complete, the task moves to **Review**. This workflow goes a long way toward producing usable results; in the author's own testing, teams have collaborated continuously for as long as 48 hours to complete complex tasks. Here is a simple example:

View full repository on GitHub →