diff --git a/.gitignore b/.gitignore
index a84a2fa..65da816 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,2 +1,14 @@
 CLAUDE.md
 blueprint.md
+
+# Server
+server/.env
+server/__pycache__/
+server/app/__pycache__/
+server/.venv/
+
+# Client
+client/node_modules/
+client/dist/
+client/android/
+client/ios/
diff --git a/server/.env.example b/server/.env.example
new file mode 100644
index 0000000..6665018
--- /dev/null
+++ b/server/.env.example
@@ -0,0 +1,12 @@
+# Obbligatorie
+NVIDIA_API_KEY=your_nvidia_api_key_here
+APP_PASSWORD=your_shared_password_here
+
+# Opzionali — i default sono in app/config.py
+# NVIDIA_BASE_URL=https://integrate.api.nvidia.com/v1
+# NVIDIA_MODEL=nvidia/nemotron-3-ultra-550b-a55b
+# NVIDIA_TIMEOUT_SECONDS=120
+# NVIDIA_MAX_TOKENS=16384
+# NVIDIA_TEMPERATURE=1.0
+# NVIDIA_TOP_P=0.95
+# NVIDIA_REASONING_BUDGET=2048
diff --git a/server/Dockerfile b/server/Dockerfile
new file mode 100644
index 0000000..d086f5c
--- /dev/null
+++ b/server/Dockerfile
@@ -0,0 +1,7 @@
+FROM python:3.11-slim
+WORKDIR /app
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+COPY app/ ./app/
+EXPOSE 8000
+CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000"]
diff --git a/server/app/__init__.py b/server/app/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/server/app/config.py b/server/app/config.py
new file mode 100644
index 0000000..880c04d
--- /dev/null
+++ b/server/app/config.py
@@ -0,0 +1,18 @@
+from pydantic_settings import BaseSettings
+
+
+class Settings(BaseSettings):
+    nvidia_api_key: str
+    app_password: str
+    nvidia_base_url: str = "https://integrate.api.nvidia.com/v1"
+    nvidia_model: str = "nvidia/nemotron-3-ultra-550b-a55b"
+    nvidia_timeout_seconds: int = 120
+    nvidia_max_tokens: int = 16384
+    nvidia_temperature: float = 1.0
+    nvidia_top_p: float = 0.95
+    nvidia_reasoning_budget: int = 2048
+
+    model_config = {"env_file": ".env"}
+
+
+settings = Settings()
diff --git a/server/app/main.py b/server/app/main.py
new file mode 100644
index 0000000..2c9534c
--- /dev/null
+++ b/server/app/main.py
@@ -0,0 +1,110 @@
+import json
+import logging
+from contextlib import asynccontextmanager
+
+from fastapi import Depends, FastAPI, HTTPException, Header
+from fastapi.middleware.cors import CORSMiddleware
+from openai import AsyncOpenAI
+
+from .config import settings
+from .models import RecipeRequest, RecipeResponse
+from .prompt import SYSTEM_PROMPT
+
+logger = logging.getLogger("uvicorn")
+
+_client: AsyncOpenAI | None = None
+
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    global _client
+    _client = AsyncOpenAI(
+        api_key=settings.nvidia_api_key,
+        base_url=settings.nvidia_base_url,
+        timeout=float(settings.nvidia_timeout_seconds),
+    )
+    yield
+    await _client.close()
+
+
+app = FastAPI(title="Recipe AI Server", lifespan=lifespan)
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+
+def _check_password(x_app_password: str = Header(...)):
+    if x_app_password != settings.app_password:
+        raise HTTPException(status_code=401, detail="Password non valida")
+
+
+@app.get("/health")
+async def health():
+    return {"status": "ok", "provider": "nvidia", "model": settings.nvidia_model}
+
+
+@app.post("/auth/test", dependencies=[Depends(_check_password)])
+async def auth_test():
+    return {"status": "ok"}
+
+
+@app.post("/recipes/generate", response_model=RecipeResponse, dependencies=[Depends(_check_password)])
+async def generate_recipes(request: RecipeRequest):
+    user_message = _build_user_message(request)
+
+    for attempt in range(3):
+        try:
+            stream = await _client.chat.completions.create(
+                model=settings.nvidia_model,
+                messages=[
+                    {"role": "system", "content": SYSTEM_PROMPT},
+                    {"role": "user", "content": user_message},
+                ],
+                temperature=settings.nvidia_temperature,
+                top_p=settings.nvidia_top_p,
+                max_tokens=settings.nvidia_max_tokens,
+                extra_body={
+                    "chat_template_kwargs": {"enable_thinking": True},
+                    "reasoning_budget": settings.nvidia_reasoning_budget,
+                },
+                stream=True,
+            )
+            content = ""
+            async for chunk in stream:
+                if not chunk.choices:
+                    continue
+                delta = chunk.choices[0].delta
+                if delta.content is not None:
+                    content += delta.content
+
+            data = json.loads(content.strip())
+            return RecipeResponse(**data)
+        except json.JSONDecodeError:
+            if attempt == 2:
+                raise HTTPException(status_code=502, detail="Il modello non ha prodotto JSON valido dopo 3 tentativi")
+            logger.warning("Tentativo %d: JSON non valido, retry...", attempt + 1)
+        except Exception as e:
+            raise HTTPException(status_code=502, detail=str(e))
+
+
+def _build_user_message(request: RecipeRequest) -> str:
+    ingredients_str = "\n".join(
+        f"- {ing.name}: {ing.quantity} {ing.unit}" for ing in request.ingredients
+    )
+    p = request.preferences
+    return (
+        f"Ingredienti disponibili:\n{ingredients_str}\n\n"
+        f"Preferenze:\n"
+        f"- Porzioni: {p.servings}\n"
+        f"- Tempo massimo: {p.maxTimeMinutes} minuti\n"
+        f"- Difficoltà: {p.difficulty}\n"
+        f"- Dieta: {p.diet}\n"
+        f"- Allergie: {', '.join(p.allergies) or 'nessuna'}\n"
+        f"- Alimenti da evitare: {', '.join(p.avoid) or 'nessuno'}\n"
+        f"- Strumenti disponibili: {', '.join(p.tools) or 'standard'}\n\n"
+        "Genera 3 ricette in JSON."
+    )
diff --git a/server/app/models.py b/server/app/models.py
new file mode 100644
index 0000000..234b822
--- /dev/null
+++ b/server/app/models.py
@@ -0,0 +1,38 @@
+from pydantic import BaseModel, Field
+
+
+class Ingredient(BaseModel):
+    name: str = Field(..., max_length=100)
+    quantity: float = Field(..., gt=0)
+    unit: str = Field(..., max_length=50)
+
+
+class Preferences(BaseModel):
+    servings: int = Field(2, ge=1, le=20)
+    maxTimeMinutes: int = Field(60, ge=5, le=300)
+    difficulty: str = Field("facile", max_length=50)
+    diet: str = Field("nessuna", max_length=100)
+    allergies: list[str] = Field(default_factory=list)
+    avoid: list[str] = Field(default_factory=list)
+    tools: list[str] = Field(default_factory=list)
+
+
+class RecipeRequest(BaseModel):
+    ingredients: list[Ingredient] = Field(..., min_length=1, max_length=30)
+    preferences: Preferences = Field(default_factory=Preferences)
+
+
+class Recipe(BaseModel):
+    title: str
+    servings: int
+    timeMinutes: int
+    difficulty: str
+    whyThisRecipe: str
+    usedIngredients: list[str]
+    optionalMissingIngredients: list[str]
+    steps: list[str]
+    notes: list[str]
+
+
+class RecipeResponse(BaseModel):
+    recipes: list[Recipe]
diff --git a/server/app/prompt.py b/server/app/prompt.py
new file mode 100644
index 0000000..1e9cac1
--- /dev/null
+++ b/server/app/prompt.py
@@ -0,0 +1,27 @@
+SYSTEM_PROMPT = """Sei un assistente cuoco esperto. Il tuo compito è generare esattamente 3 ricette in JSON valido.
+
+Regole:
+- Usa prima gli ingredienti forniti dall'utente, nelle quantità indicate.
+- Non inventare ingredienti principali non presenti nella lista.
+- Rispetta rigorosamente allergie e alimenti da evitare.
+- Adatta le ricette alle quantità indicate e al numero di porzioni.
+- Rispetta il tempo massimo di preparazione e il livello di difficoltà.
+- Produci passaggi operativi concreti e dettagliati.
+- Restituisci SOLO JSON valido, senza testo aggiuntivo, senza markdown, senza commenti.
+
+Formato risposta obbligatorio:
+{
+  "recipes": [
+    {
+      "title": "...",
+      "servings": 2,
+      "timeMinutes": 25,
+      "difficulty": "facile",
+      "whyThisRecipe": "...",
+      "usedIngredients": ["250 g pasta", "2 zucchine"],
+      "optionalMissingIngredients": ["parmigiano"],
+      "steps": ["Passo 1...", "Passo 2..."],
+      "notes": ["Nota opzionale..."]
+    }
+  ]
+}"""
diff --git a/server/docker-compose.yml b/server/docker-compose.yml
new file mode 100644
index 0000000..9838ae4
--- /dev/null
+++ b/server/docker-compose.yml
@@ -0,0 +1,8 @@
+services:
+  server:
+    build: .
+    ports:
+      - "8000:8000"
+    env_file:
+      - .env
+    restart: unless-stopped
diff --git a/server/requirements.txt b/server/requirements.txt
new file mode 100644
index 0000000..e817ddc
--- /dev/null
+++ b/server/requirements.txt
@@ -0,0 +1,6 @@
+fastapi>=0.111.0
+uvicorn[standard]>=0.29.0
+pydantic>=2.7.0
+pydantic-settings>=2.3.0
+openai>=1.30.0
+python-dotenv>=1.0.0