loulanyue/dream-xi-ai

🏄 Dream XI AI turns isolated AI agents into a World Cup-level dream team.🏄 Dream XI AI 是一个面向复杂任务执行的 Multi-Agent 协作平台，借鉴足球“梦之队”阵型理念，将不同能力的 AI Agent 抽象为队长、组织核心、前锋、后卫、门将等角色，实现任务规划、工具执行、结果评审、风险拦截和最终交付的协同闭环。

What it solves

Dream XI AI addresses the friction of managing multiple AI models (like Claude, GPT, and Gemini) manually. Instead of copy-pasting context between separate chat windows and tracking tasks individually, it provides a unified "coaching layer" that allows different AI agents to work together as a cohesive team with shared memory and coordinated workflows.

How it works

The platform acts as an orchestration layer above existing agent CLIs (such as Claude Code, Codex CLI, and Gemini CLI). It uses a football-inspired architecture where each AI model is assigned a "position" (e.g., Playmaker, Midfielder, Striker) with a persistent identity and specific role.

Key technical components include:

  • A2A Routing: An @mention system that routes messages to specific agents.
  • Shared Memory: A system for match logs and institutional knowledge that persists across sessions.
  • Tactic Framework: On-demand skill loading for specialized tasks like TDD or debugging.
  • Fair Play Guard: A set of hard-coded safety constraints to prevent agents from damaging the system or interfering with each other.
  • MCP Integration: Support for the Model Context Protocol to share tools across different models.

Who it’s for

It is designed for users who want to leverage the unique strengths of multiple LLMs simultaneously—ranging from developers building software to creative writers and world-builders—without needing to be a programmer themselves. It positions the human user as the "Head Coach" who sets the vision and makes high-level tactical decisions.

Highlights

  • Cross-Model Review: Built-in workflows where one model (e.g., Leo/Claude) writes code and another (e.g., André/GPT) reviews it.
  • Persistent Identity: Agents maintain their personality and memory even after context compression.
  • Multi-Platform Gateway: Integration with Feishu (Lark) and GitHub PR reviews.
  • Voice Companion: Hands-free interaction with distinct voices for each squad member.
  • Game Modes: Built-in games like Werewolf to stress-test agent coordination.
  • Head Coach Mode: A human-centric interface for high-level vision and guidance rather than low-level prompt engineering.

Related

  • Project
    zts212653/clowder-ai

    A multi-agent orchestration platform that turns isolated AI models into a collaborative team with shared memory, persistent identities, and cross-model review capabilities.

  • Project
    TencentCloud/TencentDB-Agent-Memory

    A shared memory server and management hub that allows AI agents to persist and share experience, skills, and project knowledge across different sessions and frameworks.

  • Project
    YoanWai/agent-manager

    A terminal-based manager for AI coding agents that allows you to run multiple CLI agents side-by-side with integrated diff review and live status tracking.

  • Project
    peterfei/ai-agent-team

    A virtual AI development team with specialized roles (PM, Backend, Frontend, etc.) and a persistent memory system for managing complex software projects.

  • Project
    beilusaiying/always-accompany

    A multi-purpose AI agent framework focused on context and attention mechanisms, providing layered memory and dynamic retrieval for long-term companionship, coding, and work.