Add: Sampling Example to README

README.md

@@ -12,6 +12,7 @@
   - [Tools](#tools)
   - [Prompts](#prompts)
   - [Completions](#completions)
+  - [Sampling](#sampling)
 - [Running Your Server](#running-your-server)
   - [stdio](#stdio)
   - [Streamable HTTP](#streamable-http)
@@ -382,6 +383,37 @@ import { getDisplayName } from "@modelcontextprotocol/sdk/shared/metadataUtils.j
 const displayName = getDisplayName(tool);
 ```
 
+### Sampling
+
+MCP servers can also request MCP client LLMs for responses. Below is an example of a sampling request sent just after connecting to the Client
+
+```typescript
+// Result sent back from LLM follow the CreateMessageSchema
+import {CreateMessageResult} from "@modelcontextprotocol/sdk/types.js";
+
+// Async Function to send a sampling request to the LLM at top-level
+async function samplingExample(server: McpServer): Promise<CreateMessageResult> {	
+	const samplingText = "Text prompt to send to LLM";
+	const result = await McpServer.server.createMessage(
+		{
+			messages : [{
+				role: "user",
+				content: {
+					text: samplingText,
+				type: "text"
+				}
+			}],
+			maxTokens: 1000
+		}
+	);
+	return result;
+}
+
+// Sampling request just after connecting to MCP Client
+server.connect(transport);
+samplingExample(server);
+```
+
 ## Running Your Server
 
 MCP servers in TypeScript need to be connected to a transport to communicate with clients. How you start the server depends on the choice of transport: