docs(mcp): add nuxt ui agent as a mcp tool

docs/content/docs/1.getting-started/7.ai/1.mcp.md

@@ -54,6 +54,14 @@ The Nuxt UI MCP server provides the following tools organized by category:
 
 - **`get_migration_guide`**: Retrieves version-specific migration guides and upgrade instructions
 
+### Assistant Tools
+
+- **`ask_nuxt_ui_agent`**: Ask the Nuxt UI assistant for a concise direct answer (returns only the final content, no intermediate steps).
+
+::note{icon="i-lucide-info"}
+Disabled when using the Agent from the Nuxt UI website.
+::
+
 ## Available Prompts
 
 The Nuxt UI MCP server provides guided prompts for common workflows:

docs/server/api/search.ts

@@ -1,10 +1,10 @@
 import { StreamableHTTPClientTransport } from '@modelcontextprotocol/sdk/client/streamableHttp.js'
-import { streamText, convertToModelMessages, stepCountIs } from 'ai'
+import { streamText, convertToModelMessages, stepCountIs, generateText } from 'ai'
 import { experimental_createMCPClient } from '@ai-sdk/mcp'
 import { gateway } from '@ai-sdk/gateway'
 
 export default defineEventHandler(async (event) => {
-  const { messages } = await readBody(event)
+  const { messages, stream = true } = await readBody(event)
 
   const httpTransport = new StreamableHTTPClientTransport(
     new URL(import.meta.dev ? 'http://localhost:3000/mcp' : 'https://ui.nuxt.com/mcp')
@@ -14,9 +14,12 @@ export default defineEventHandler(async (event) => {
   })
   const tools = await httpClient.tools()
 
-  return streamText({
+  // Remove the ask_nuxt_ui_agent tool to avoid infinite loops
+  delete tools['ask_nuxt_ui_agent']
+
+  const options = {
     model: gateway('anthropic/claude-sonnet-4.5'),
-    maxOutputTokens: 10000,
+    maxOutputTokens: 10_000,
     system: `You are a helpful assistant for Nuxt UI, a UI library for Nuxt and Vue. Use your knowledge base tools to search for relevant information before answering questions.
 
 Guidelines:
@@ -40,7 +43,23 @@ Guidelines:
 - Format responses in a conversational way, not as documentation sections.
     `,
     messages: convertToModelMessages(messages),
-    stopWhen: stepCountIs(6),
+    stopWhen: stepCountIs(6)
+  }
+
+  // If not streaming, it's called from the MCP route as a tool.
+  if (stream === false) {
+    const { text } = await generateText({
+      ...options,
+      tools
+    })
+
+    await httpClient.close()
+
+    return text
+  }
+
+  return streamText({
+    ...options,
     tools,
     onFinish: async () => {
       await httpClient.close()

docs/server/routes/mcp.ts

@@ -5,6 +5,7 @@
 import { z } from 'zod/v3'
 import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'
 import { StreamableHTTPServerTransport } from '@modelcontextprotocol/sdk/server/streamableHttp.js'
+import type { UIMessage } from 'ai'
 
 function createServer() {
   const server = new McpServer({
@@ -413,6 +414,36 @@ function createServer() {
     }
   )
 
+  server.registerTool(
+    'ask_nuxt_ui_agent',
+    {
+      title: 'Ask Nuxt UI Agent',
+      description: 'Asks the Nuxt UI agent a question',
+      inputSchema: {
+        // @ts-expect-error - need to wait for support for zod 4, this works correctly just a type mismatch from zod 3 to zod 4 (https://github.com/modelcontextprotocol/typescript-sdk/pull/869)
+        query: z.string().describe('The question to ask the agent')
+      }
+    },
+    async (params: { query: string }) => {
+      const result = await $fetch<string>('/api/search', {
+        method: 'POST',
+        timeout: 60_000,
+        body: {
+          stream: false,
+          messages: [{
+            role: 'user',
+            parts: [{ type: 'text', text: params.query }]
+          }] satisfies Array<Omit<UIMessage, 'id'>> }
+      })
+
+      console.log(result)
+
+      return {
+        content: [{ type: 'text' as const, text: result }]
+      }
+    }
+  )
+
   return server
 }