diff --git a/designer-demo/src/composable/http/index.js b/designer-demo/src/composable/http/index.js
index f11538395a..633ccb1bc1 100644
--- a/designer-demo/src/composable/http/index.js
+++ b/designer-demo/src/composable/http/index.js
@@ -46,7 +46,7 @@ const preResponse = (res) => {
     return Promise.reject(res.data.error)
   }
 
-  return res.data?.data
+  return res.data?.data || res.data
 }
 
 const openLogin = () => {
diff --git a/packages/plugins/robot/src/js/robotSetting.ts b/packages/plugins/robot/src/js/robotSetting.ts
index ddbdf02343..999181a602 100644
--- a/packages/plugins/robot/src/js/robotSetting.ts
+++ b/packages/plugins/robot/src/js/robotSetting.ts
@@ -19,7 +19,7 @@ const DEFAULT_MODELS = [{ label: 'DeepSeek：DeepSeek-V3', value: 'deepseek-chat
 
 export const getAIModelOptions = () => {
   const aiRobotOptions = getOptions(meta.id)?.customCompatibleAIModels || []
-  return [...DEFAULT_MODELS, ...aiRobotOptions]
+  return aiRobotOptions.length ? aiRobotOptions : DEFAULT_MODELS
 }
 
 // 这里存放的是aichat的响应式数据
diff --git a/packages/plugins/robot/src/mcp/MarkdownRenderer.vue b/packages/plugins/robot/src/mcp/MarkdownRenderer.vue
index 6578667985..6fc7ba4ab6 100644
--- a/packages/plugins/robot/src/mcp/MarkdownRenderer.vue
+++ b/packages/plugins/robot/src/mcp/MarkdownRenderer.vue
@@ -26,7 +26,7 @@ const defaultMarkdownItOptions = {
 }
 
 const props = defineProps<{
-  type: 'markdown' | 'text'
+  type?: 'markdown' | 'text'
   content: string
   options?: MarkdownItOptions
 }>()
diff --git a/packages/plugins/robot/src/mcp/utils.ts b/packages/plugins/robot/src/mcp/utils.ts
index 5ef759a8d2..8b785bca98 100644
--- a/packages/plugins/robot/src/mcp/utils.ts
+++ b/packages/plugins/robot/src/mcp/utils.ts
@@ -2,6 +2,7 @@ import { toRaw } from 'vue'
 import useMcpServer from './useMcp'
 import type { LLMMessage, RobotMessage } from './types'
 import type { LLMRequestBody, LLMResponse, ReponseToolCall, RequestOptions, RequestTool } from './types'
+import { META_SERVICE, getMetaApi } from '@opentiny/tiny-engine-meta-register'
 
 let requestOptions: RequestOptions = {}
 
@@ -14,13 +15,11 @@ const fetchLLM = async (messages: LLMMessage[], tools: RequestTool[], options: R
   if (tools.length > 0) {
     bodyObj.tools = toRaw(tools)
   }
-  return fetch(options?.url || '/app-center/api/chat/completions', {
-    method: 'POST',
+  return getMetaApi(META_SERVICE.Http).post(options?.url || '/app-center/api/chat/completions', bodyObj, {
     headers: {
       'Content-Type': 'application/json',
       ...options?.headers
-    },
-    body: JSON.stringify(bodyObj)
+    }
   })
 }
 
@@ -82,7 +81,7 @@ const handleToolCall = async (
         result: toolCallResult.content
       }
     }
-    const newResp = await fetchLLM(toolMessages, tools).then((res) => res.json())
+    const newResp = await fetchLLM(toolMessages, tools)
     const hasToolCall = newResp.choices[0].message.tool_calls?.length > 0
     if (hasToolCall) {
       await handleToolCall(newResp, tools, messages, toolMessages)
@@ -103,7 +102,7 @@ export const sendMcpRequest = async (messages: LLMMessage[], options: RequestOpt
   }
   const tools = await useMcpServer().getLLMTools()
   requestOptions = options
-  const res = await fetchLLM(messages.slice(0, -1), tools, options).then((res) => res.json())
+  const res = await fetchLLM(messages.slice(0, -1), tools, options)
   const hasToolCall = res.choices[0].message.tool_calls?.length > 0
   if (hasToolCall) {
     await handleToolCall(res, tools, messages)