diff --git a/backend/app/schemas/worker.py b/backend/app/schemas/worker.py
new file mode 100644
index 0000000..bdf1a91
--- /dev/null
+++ b/backend/app/schemas/worker.py
@@ -0,0 +1,72 @@
+"""AI Worker schemas."""
+from pydantic import BaseModel, Field
+from typing import Optional
+from datetime import datetime
+
+
+class WorkerCreate(BaseModel):
+    name: str = Field(..., min_length=1, max_length=200)
+    description: str = ""
+    provider: str = "openai"
+    model_name: str = "gpt-4o-mini"
+    api_key: Optional[str] = None  # If None, uses platform default
+    base_url: Optional[str] = None
+    temperature: float = 0.7
+    system_prompt: str = "You are a helpful AI assistant."
+    tools: str = "[]"
+    max_cost_per_task_cents: Optional[int] = None
+    max_cost_per_month_cents: Optional[int] = None
+    available_hours: str = '{"start": "00:00", "end": "23:59"}'
+    max_concurrent_tasks: int = 1
+
+
+class WorkerUpdate(BaseModel):
+    name: Optional[str] = None
+    description: Optional[str] = None
+    provider: Optional[str] = None
+    model_name: Optional[str] = None
+    api_key: Optional[str] = None
+    base_url: Optional[str] = None
+    temperature: Optional[float] = None
+    system_prompt: Optional[str] = None
+    tools: Optional[str] = None
+    max_cost_per_task_cents: Optional[int] = None
+    max_cost_per_month_cents: Optional[int] = None
+    is_active: Optional[bool] = None
+    available_hours: Optional[str] = None
+    max_concurrent_tasks: Optional[int] = None
+    status: Optional[str] = None
+
+
+class WorkerResponse(BaseModel):
+    id: int
+    name: str
+    description: str
+    provider: str
+    model_name: str
+    base_url: Optional[str] = None
+    temperature: float
+    system_prompt: str
+    tools: str
+    max_cost_per_task_cents: Optional[int] = None
+    max_cost_per_month_cents: Optional[int] = None
+    status: str
+    is_active: bool
+    current_task_count: int
+    max_concurrent_tasks: int
+    tenant_id: int
+    created_at: datetime
+    updated_at: datetime
+
+    model_config = {"from_attributes": True}
+
+
+class WorkerTestRequest(BaseModel):
+    test_prompt: str = "Hello, respond with 'OK' if you can hear me."
+
+
+class WorkerTestResult(BaseModel):
+    success: bool
+    response: str = ""
+    latency_ms: int = 0
+    error: Optional[str] = None