** 多代理系统,8名AI代理作为团队工作。 ** 建于OpenClaw上. 被"破坏"所启发.

□ Canonical source(英语:Canonical source) 出自 □ Canonical source(英语:Canonical source).

该项目由Aleksei Ulianov / Sprut AI维护. 原始储存库: https://github.com/AlekseiUL/heisenberg-team

如果您发现此工程有镜像,重新包装, 或重新分配到其他地方, 请检查此寄存器是否为真相来源 。

□ 归属

在适用许可证允许的情况下,如果重新使用,叉子,修改,包装,或发布本作,保留原版权和许可证通知并链接回犬科库.

第2版增加了更好的直接信息行为,耐久记忆规则,并更新了OpenClaw操作.

– – . . .

□ 目录

• 这是什么?
• 为什么?
• 这有什么不同?
• 建筑
• 探员
• 快开始
• 技能(34) 项目结构
• 实例
• 文档
• 贡献
• 执照

– – . . .

□ 要求

• 节点.js v24 LTS
• OpenClaw([T1])
• 至少一个LLM供应商(Anthropic、OpenAI、OpenAI Codex、Google、DeepSeek或本地模型)的认证
• Telegram bot 令牌(可选,通过 @BotFather通知)

• 系统要求

🧪 Heisenberg Team

A multi-agent system with 8 AI agents working as a team. Built on OpenClaw. Inspired by Breaking Bad.

Canonical source

This project is maintained by Aleksei Ulianov / Sprut_AI. Original repository: https://github.com/AlekseiUL/heisenberg-team

If you found this project mirrored, repackaged, or redistributed elsewhere, check this repository as the source of truth.

Attribution

Where permitted by the applicable license, if you reuse, fork, modify, package, or publish this work, keep the original copyright and license notice and link back to the canonical repository.

Release 2 adds better direct-message behavior, durable-memory rules, and updated OpenClaw operations.

---

Table of Contents

• What is this?
• Why?
• How is this different?
• Architecture
• Agents
• Quick Start
• Skills (34)
• Project Structure
• Examples
• Documentation
• Contributing
• License

---

Requirements

• Node.js v24 LTS
• OpenClaw (npm install -g openclaw)
• Auth for at least one LLM provider (Anthropic, OpenAI, OpenAI Codex, Google, DeepSeek, or local models)
• Telegram bot token (optional, for notifications via @BotFather)

System Requirements

What is this?

A production-ready template for running a team of AI agents that communicate, delegate tasks, and deliver results. Each agent has a defined role, personality, and set of skills.

This is not a framework. This is a working system you can clone, configure, and run.

Who is this for?

Use this repository if you want a self-hosted OpenClaw team that feels like an operating crew, not a single chatbot:

• founders and operators who want a Telegram-accessible AI command center;
• builders experimenting with multi-agent delegation and durable task boards;
• AI automation consultants who need a concrete reference implementation;
• OpenClaw users who want memory, recovery, skills, and specialist roles already wired together.