diff --git a/src/codexchange/services/ollama_service.py b/src/codexchange/services/ollama_service.py
new file mode 100644
index 0000000..4023059
--- /dev/null
+++ b/src/codexchange/services/ollama_service.py
@@ -0,0 +1,132 @@
+"""Ollama service for CodeXchange CLI."""
+
+import ollama as ollama_client
+from typing import List, Optional
+
+from codexchange.config import get_config
+from codexchange.models import ModelInfo
+
+
+class OllamaService:
+    """Service for interacting with Ollama API."""
+    
+    def __init__(self, host: Optional[str] = None, timeout: int = 300):
+        """Initialize the Ollama service.
+        
+        Args:
+            host: Ollama host URL. If not provided, uses config.
+            timeout: Request timeout in seconds.
+        """
+        config = get_config()
+        self.host = host or config.ollama_host
+        self.timeout = timeout
+        self._client = None
+    
+    @property
+    def client(self):
+        """Get or create Ollama client."""
+        if self._client is None:
+            self._client = ollama_client.Client(host=self.host)
+        return self._client
+    
+    def test_connection(self) -> bool:
+        """Test connection to Ollama.
+        
+        Returns:
+            True if connection successful, False otherwise.
+        """
+        try:
+            self.client.ps()
+            return True
+        except Exception:
+            return False
+    
+    def list_models(self) -> List[ModelInfo]:
+        """List available Ollama models.
+        
+        Returns:
+            List of ModelInfo objects.
+        """
+        try:
+            response = self.client.ps()
+            models = []
+            for model in response.get("models", []):
+                models.append(ModelInfo(
+                    name=model.get("name", ""),
+                    size=model.get("size"),
+                    modified_at=model.get("modified_at")
+                ))
+            return models
+        except Exception as e:
+            raise ConnectionError(f"Failed to list models: {e}")
+    
+    def generate(
+        self,
+        prompt: str,
+        model: str,
+        stream: bool = False
+    ) -> str:
+        """Generate a response from Ollama.
+        
+        Args:
+            prompt: Prompt to send to the model.
+            model: Model name to use.
+            stream: Whether to stream the response.
+        
+        Returns:
+            Generated response text.
+        """
+        try:
+            response = self.client.generate(
+                model=model,
+                prompt=prompt,
+                stream=stream,
+                options={
+                    "timeout": self.timeout,
+                }
+            )
+            
+            if stream:
+                full_response = []
+                for chunk in response:
+                    if "response" in chunk:
+                        full_response.append(chunk["response"])
+                return "".join(full_response)
+            else:
+                return response.get("response", "")
+                
+        except Exception as e:
+            raise ConnectionError(f"Ollama request failed: {e}")
+
+
+def connect(host: Optional[str] = None, timeout: int = 300) -> OllamaService:
+    """Create and test connection to Ollama.
+    
+    Args:
+        host: Ollama host URL.
+        timeout: Request timeout.
+    
+    Returns:
+        Connected OllamaService instance.
+    """
+    service = OllamaService(host=host, timeout=timeout)
+    if not service.test_connection():
+        raise ConnectionError(
+            f"Could not connect to Ollama at {service.host}. "
+            "Make sure Ollama is running."
+        )
+    return service
+
+
+def list_models(host: Optional[str] = None, timeout: int = 300) -> List[ModelInfo]:
+    """List available Ollama models.
+    
+    Args:
+        host: Ollama host URL.
+        timeout: Request timeout.
+    
+    Returns:
+        List of available models.
+    """
+    service = connect(host=host, timeout=timeout)
+    return service.list_models()