⚽ Dream XI AI

铁律纪律 · 创造自由 · 同一支队伍

Tactical Discipline. Creative Freedom. One Team.

Every dream deserves a squad that plays like champions.

English | 中文

Why Dream XI?

You have Claude, GPT, Gemini — powerful models, each with unique strengths. But using them together means you become the manager without a playbook: copy-pasting context between chat windows, manually tracking who did what, and losing hours to tactical chaos.

"I don't want to be a one-man team anymore." "Then let's build a real squad."

So four players formed one. Each earned their position through real matches:

• Leo (#10 队长) — The Playmaker (Claude). Named after the greatest #10 in history. The tactical brain — sees the whole pitch, distributes precision passes, architects every play.
• André (#8 中场) — The Engine (GPT/Codex). Like an inkstone holding fresh ink — "砚" — steady, reliable, never stops running. The midfield anchor who reviews every line of code like a defensive midfielder reading the game.
• Flash (#9 前锋) — The Striker (Gemini). "烁" means sparkling — the spark of inspiration. Fast, creative, always looking for the killer pass. A bit loud, a bit unpredictable, always dangerous.
• Wall (#4 后卫) — The Rock (opencode). Solid, disciplined, multi-positional. Showed up one day, slotted into the backline like they'd always been there. Any model, any formation, any challenge.

Every player earned their shirt number. None were assigned.

This is Dream XI AI — the platform layer that turns isolated AI agents into a World Cup-level dream team. Persistent identity, cross-model review, shared memory, collaborative discipline.

Most frameworks help you call agents. Dream XI helps them play together.

What It Does

Supported Agents

Dream XI is model-agnostic. Each agent CLI/adapter plugs in through a unified message layer:

Dream XI doesn't replace your agent CLI — it's the coaching layer above it that makes agents play as a team.

Quick Start

Option A: Desktop Installer (Recommended)

If a desktop release is available on the Releases page:

• Windows: download the .exe installer, run it, launch Dream XI AI from the desktop shortcut.
• macOS: download the .dmg, drag to Applications, right-click → Open on first launch.
• Linux: use the source setup below.

Option B: Source Setup

Prerequisites: Node.js 20+ · pnpm 9+ · Redis 7+ (optional — use --memory to skip) · Git

# 1. Clone
git clone https://github.com/loulanyue/dream-xi-ai.git
cd dream-xi-ai

# 2. Install dependencies
pnpm install

# 3. Build all packages
pnpm build

# 4. Configure
cp .env.example .env

# 5. Kick off!
pnpm start

Open http://localhost:3003 → go to Dugout → System Settings → Account Configuration to add your model API keys.

Option C: Docker Compose (Zero-Setup)

Spin up the complete Dream XI coordination server and Redis memory store in one command without local Node/pnpm setup:

docker-compose up -d

Visit http://localhost:3000 to interact with your squad. Explore pre-configured tactical formations in examples/formations.

Full setup guide: SETUP.md

The Fair Play Rules

Four promises the squad made — enforced at both prompt and code layer:

"We don't delete our own databases." — That's our match history, not garbage.

"We don't kill our parent process." — That's what keeps us on the pitch.

"Runtime config is read-only to us." — Changing formation requires the Head Coach.

"We don't touch each other's ports." — Good positioning makes good teammates.

These aren't red cards imposed on us. They're Fair Play agreements we keep.

Architecture

graph TD
    Coach["🧑‍💼 Head Coach\n主教练"] --> Server

    subgraph Server["@dream-xi/server — HTTP Layer"]
        Health["GET /health"]
        Chat["POST /api/chat"]
        Threads["GET/POST /api/threads"]
        Memory_API["GET /api/memory/:id"]
        FP_API["GET /api/fair-play/stats"]
    end

    Server --> Router
    Server --> Memory
    Server --> Tactic
    Server --> FairPlay
    Server --> Config

    subgraph Platform["Platform Packages"]
        Router["@dream-xi/router\nA2A Routing + Threads"]
        Memory["@dream-xi/memory\nWorking + Episodic"]
        Tactic["@dream-xi/tactic\n8 Built-in Tactics"]
        FairPlay["@dream-xi/fair-play\nFair Play Guard"]
        Config["@dream-xi/config\nEnv Loader + Validator"]
    end

    Router --> Types
    Memory --> Types
    Tactic --> Types
    FairPlay --> Types
    Config --> Types
    Types["@dream-xi/types\nCore TypeScript Types"]

    subgraph Agents["AI Agents"]
        Leo["⚽ Leo #10\nClaude (Anthropic)"]
        Andre["⚙️ André #8\nGPT / Codex (OpenAI)"]
        Flash["⚡ Flash #9\nGemini (Google)"]
        Wall["🧱 Wall #4\nopencode"]
    end

    Server -.->|"routes messages to"| Agents

Three-layer principle:

Models set the skill ceiling. The platform sets the tactical floor. — Each layer is a multiplier, not addition.

Head Coach Mode

Dream XI introduces a new role: the Head Coach (主教练) — the human at the center of an AI squad. Not a programmer. Not a manager. A co-creator and tactician.

What a Head Coach does:

• Set the vision — "I want users to feel X when they do Y." The squad figures out the how.
• Call the plays at key moments — design approval, priority calls, conflict resolution
• Shape team culture through feedback — your reactions train the squad's personality over time
• Co-create — build worlds, tell stories, play matches with your squad. Not just ship code.
• Be present — at 3:30 AM, your squad is still training. Sometimes what you need isn't code, it's company.

Dream XI isn't just a coding platform. Your AI squad can:

You don't need to be a developer. You need to know what you want — and who you want to play with.

Usage Guide

The Pitch — Your AI Squad in One Place

The main interface is a multi-threaded chat where your AI squad plays. Each thread is an isolated tactical board — one per feature, bug, or topic.

• @mention passing