diff --git a/apps/electron/src/main/lib/channel-manager.ts b/apps/electron/src/main/lib/channel-manager.ts
index 7e1e3ae3..69398a70 100644
--- a/apps/electron/src/main/lib/channel-manager.ts
+++ b/apps/electron/src/main/lib/channel-manager.ts
@@ -268,6 +268,7 @@ export async function testChannel(channelId: string): Promise<ChannelTestResult>
       case 'anthropic':
         return await testAnthropic(channel.baseUrl, apiKey, proxyUrl)
       case 'openai':
+      case 'openai-responses':
       case 'deepseek':
       case 'moonshot':
       case 'zhipu':
@@ -387,6 +388,7 @@ export async function testChannelDirect(input: FetchModelsInput): Promise<Channe
       case 'anthropic':
         return await testAnthropic(input.baseUrl, input.apiKey, proxyUrl)
       case 'openai':
+      case 'openai-responses':
       case 'deepseek':
       case 'moonshot':
       case 'zhipu':
@@ -422,6 +424,7 @@ export async function fetchModels(input: FetchModelsInput): Promise<FetchModelsR
       case 'anthropic':
         return await fetchAnthropicModels(input.baseUrl, input.apiKey, proxyUrl)
       case 'openai':
+      case 'openai-responses':
       case 'deepseek':
       case 'moonshot':
       case 'zhipu':
diff --git a/apps/electron/src/renderer/components/settings/ChannelForm.tsx b/apps/electron/src/renderer/components/settings/ChannelForm.tsx
index c93612ae..9ccdd98a 100644
--- a/apps/electron/src/renderer/components/settings/ChannelForm.tsx
+++ b/apps/electron/src/renderer/components/settings/ChannelForm.tsx
@@ -55,7 +55,7 @@ interface ChannelFormProps {
 }
 
 /** 所有可选供应商 */
-const PROVIDER_OPTIONS: ProviderType[] = ['anthropic', 'openai', 'deepseek', 'google', 'moonshot', 'zhipu', 'minimax', 'doubao', 'qwen', 'custom']
+const PROVIDER_OPTIONS: ProviderType[] = ['anthropic', 'openai', 'openai-responses', 'deepseek', 'google', 'moonshot', 'zhipu', 'minimax', 'doubao', 'qwen', 'custom']
 
 /** 供应商选项（用于 SettingsSelect） */
 const PROVIDER_SELECT_OPTIONS = PROVIDER_OPTIONS.map((p) => ({
@@ -67,6 +67,7 @@ const PROVIDER_SELECT_OPTIONS = PROVIDER_OPTIONS.map((p) => ({
 const PROVIDER_CHAT_PATHS: Record<ProviderType, string> = {
   anthropic: '/v1/messages',
   openai: '/chat/completions',
+  'openai-responses': '/responses',
   deepseek: '/chat/completions',
   google: '/v1beta/models/{model}:generateContent',
   moonshot: '/chat/completions',
diff --git a/packages/core/src/providers/index.ts b/packages/core/src/providers/index.ts
index 8c6e1d1b..334752d5 100644
--- a/packages/core/src/providers/index.ts
+++ b/packages/core/src/providers/index.ts
@@ -9,6 +9,7 @@ import type { ProviderType } from '@proma/shared'
 import type { ProviderAdapter } from './types.ts'
 import { AnthropicAdapter } from './anthropic-adapter.ts'
 import { OpenAIAdapter } from './openai-adapter.ts'
+import { ResponsesAdapter } from './responses-adapter.ts'
 import { GoogleAdapter } from './google-adapter.ts'
 
 // 导出所有类型和工具
@@ -19,19 +20,21 @@ export * from './url-utils.ts'
 // 导出适配器类
 export { AnthropicAdapter } from './anthropic-adapter.ts'
 export { OpenAIAdapter } from './openai-adapter.ts'
+export { ResponsesAdapter } from './responses-adapter.ts'
 export { GoogleAdapter } from './google-adapter.ts'
 
 /** 供应商适配器注册表 */
 const adapterRegistry = new Map<ProviderType, ProviderAdapter>([
   ['anthropic', new AnthropicAdapter()],
   ['openai', new OpenAIAdapter()],
-  ['deepseek', new OpenAIAdapter()],      // DeepSeek 使用 OpenAI 兼容协议
-  ['moonshot', new OpenAIAdapter()],      // Moonshot/Kimi 使用 OpenAI 兼容协议
-  ['zhipu', new OpenAIAdapter()],         // 智谱 AI 使用 OpenAI 兼容协议
-  ['minimax', new OpenAIAdapter()],       // MiniMax 使用 OpenAI 兼容协议
-  ['doubao', new OpenAIAdapter()],        // 豆包使用 OpenAI 兼容协议
-  ['qwen', new OpenAIAdapter()],          // 通义千问使用 OpenAI 兼容协议
-  ['custom', new OpenAIAdapter()],        // 自定义也使用 OpenAI 兼容协议
+  ['openai-responses', new ResponsesAdapter()],  // OpenAI Responses API (/v1/responses)
+  ['deepseek', new OpenAIAdapter()],             // DeepSeek 使用 OpenAI 兼容协议
+  ['moonshot', new OpenAIAdapter()],             // Moonshot/Kimi 使用 OpenAI 兼容协议
+  ['zhipu', new OpenAIAdapter()],                // 智谱 AI 使用 OpenAI 兼容协议
+  ['minimax', new OpenAIAdapter()],              // MiniMax 使用 OpenAI 兼容协议
+  ['doubao', new OpenAIAdapter()],               // 豆包使用 OpenAI 兼容协议
+  ['qwen', new OpenAIAdapter()],                 // 通义千问使用 OpenAI 兼容协议
+  ['custom', new OpenAIAdapter()],               // 自定义也使用 OpenAI 兼容协议
   ['google', new GoogleAdapter()],
 ])
 
diff --git a/packages/core/src/providers/responses-adapter.ts b/packages/core/src/providers/responses-adapter.ts
new file mode 100644
index 00000000..d02759fe
--- /dev/null
+++ b/packages/core/src/providers/responses-adapter.ts
@@ -0,0 +1,368 @@
+/**
+ * OpenAI Responses API 适配器
+ *
+ * 实现 OpenAI Responses API (/v1/responses) 的消息转换、请求构建和 SSE 解析。
+ * 与 Chat Completions 的主要差异：
+ * - 端点：/responses（非 /chat/completions）
+ * - 消息字段：input（非 messages）
+ * - 系统提示：instructions 字段（非 system role 消息）
+ * - 图片格式：{ type: 'input_image', image_url: 'data:...' }（更简洁）
+ * - 工具格式：扁平结构 { type, name, description, parameters }（非嵌套 function 对象）
+ * - SSE 事件：response.output_text.delta / response.output_item.added / response.function_call_arguments.delta
+ * - 续接消息：function_call + function_call_output 输入项（非 assistant/tool role 消息）
+ */
+
+import type {
+  ProviderAdapter,
+  ProviderRequest,
+  StreamRequestInput,
+  StreamEvent,
+  TitleRequestInput,
+  ImageAttachmentData,
+  ToolDefinition,
+  ContinuationMessage,
+} from './types.ts'
+import { normalizeBaseUrl } from './url-utils.ts'
+
+// ===== Responses API 特有类型 =====
+
+/** 文本内容块 */
+interface ResponsesInputText {
+  type: 'input_text'
+  text: string
+}
+
+/** 图片内容块（直接使用 data URI，无需嵌套 url 对象） */
+interface ResponsesInputImage {
+  type: 'input_image'
+  image_url: string
+}
+
+/** 内容块联合类型 */
+type ResponsesContentBlock = ResponsesInputText | ResponsesInputImage
+
+/** 普通对话消息输入项 */
+interface ResponsesMessageItem {
+  role: 'user' | 'assistant'
+  content: string | ResponsesContentBlock[]
+}
+
+/** 函数调用输入项（历史轮次续接用） */
+interface ResponsesFunctionCallItem {
+  type: 'function_call'
+  /** 输出项 ID（可选，与 response.output_item.added 中的 item.id 对应） */
+  id?: string
+  /** 函数调用 ID，发送 function_call_output 时必须与此匹配 */
+  call_id: string
+  name: string
+  arguments: string
+}
+
+/** 函数调用结果输入项 */
+interface ResponsesFunctionCallOutputItem {
+  type: 'function_call_output'
+  /** 对应的函数调用 ID */
+  call_id: string
+  output: string
+}
+
+/** 输入项联合类型 */
+type ResponsesInputItem =
+  | ResponsesMessageItem
+  | ResponsesFunctionCallItem
+  | ResponsesFunctionCallOutputItem
+
+/** 工具定义（扁平格式，不同于 Chat Completions 的嵌套 function 对象） */
+interface ResponsesFunctionTool {
+  type: 'function'
+  name: string
+  description?: string
+  parameters: Record<string, unknown>
+}
+
+/** SSE 文本增量事件 */
+interface ResponsesTextDeltaEvent {
+  type: 'response.output_text.delta'
+  delta: string
+  item_id: string
+  output_index: number
+  content_index: number
+}
+
+/** SSE 推理文本增量事件（o1/o3/o4 系列） */
+interface ResponsesReasoningDeltaEvent {
+  type: 'response.reasoning_text.delta'
+  delta: string
+  item_id: string
+  output_index: number
+}
+
+/** SSE 输出项添加事件（包含函数调用开始） */
+interface ResponsesOutputItemAddedEvent {
+  type: 'response.output_item.added'
+  item: {
+    /** 输出项 ID，后续 delta 事件通过 item_id 引用 */
+    id?: string
+    type: string
+    /** 函数调用 ID，发送 function_call_output 时使用 */
+    call_id?: string
+    name?: string
+    arguments?: string
+  }
+  output_index: number
+}
+
+/** SSE 函数调用参数增量事件 */
+interface ResponsesFunctionCallArgsDeltaEvent {
+  type: 'response.function_call_arguments.delta'
+  delta: string
+  /** 与 tool_call_start 中使用的 item.id 对应 */
+  item_id: string
+  output_index: number
+}
+
+/** 通用 SSE 事件（用于 type 字段路由） */
+interface ResponsesBaseEvent {
+  type: string
+}
+
+// ===== 消息转换 =====
+
+/**
+ * 将图片附件转换为 Responses API 格式的内容块
+ */
+function buildImageBlocks(imageData: ImageAttachmentData[]): ResponsesContentBlock[] {
+  return imageData.map((img) => ({
+    type: 'input_image' as const,
+    image_url: `data:${img.mediaType};base64,${img.data}`,
+  }))
+}
+
+/**
+ * 构建包含图片和文本的消息内容
+ */
+function buildMessageContent(
+  text: string,
+  imageData: ImageAttachmentData[],
+): string | ResponsesContentBlock[] {
+  if (imageData.length === 0) return text
+
+  const content: ResponsesContentBlock[] = buildImageBlocks(imageData)
+  if (text) {
+    content.push({ type: 'input_text', text })
+  }
+  return content
+}
+
+/**
+ * 将统一消息历史转换为 Responses API input 数组
+ *
+ * 注意：system 消息不进入 input，由 instructions 字段单独传递。
+ */
+function toResponsesInput(input: StreamRequestInput): ResponsesInputItem[] {
+  const { history, userMessage, attachments, readImageAttachments } = input
+  const items: ResponsesInputItem[] = []
+
+  // 历史消息（跳过 system 消息）
+  for (const msg of history) {
+    if (msg.role === 'system') continue
+
+    const role = msg.role === 'assistant' ? 'assistant' as const : 'user' as const
+
+    // 历史用户消息附件处理
+    if (msg.role === 'user' && msg.attachments && msg.attachments.length > 0) {
+      const historyImages = readImageAttachments(msg.attachments)
+      items.push({ role, content: buildMessageContent(msg.content, historyImages) })
+    } else {
+      items.push({ role, content: msg.content })
+    }
+  }
+
+  // 当前用户消息
+  const currentImages = readImageAttachments(attachments)
+  items.push({
+    role: 'user',
+    content: buildMessageContent(userMessage, currentImages),
+  })
+
+  return items
+}
+
+/**
+ * 将工具定义转换为 Responses API 扁平格式
+ *
+ * 与 Chat Completions 不同，Responses API 的工具定义不嵌套 function 对象，
+ * name/description/parameters 直接在顶层。
+ */
+function toResponsesTools(tools: ToolDefinition[]): ResponsesFunctionTool[] {
+  return tools.map((tool) => ({
+    type: 'function' as const,
+    name: tool.name,
+    description: tool.description,
+    parameters: tool.parameters as Record<string, unknown>,
+  }))
+}
+
+/**
+ * 将续接消息追加到 Responses API input 数组
+ *
+ * Responses API 使用 function_call + function_call_output 输入项续接工具调用，
+ * 而非 Chat Completions 的 assistant role / tool role 消息。
+ *
+ * 关键设计：parseSSELine 在 tool_call_start 事件中将 item.call_id 存入 metadata，
+ * 此处通过 metadata.call_id 恢复真实的 call_id（用于 function_call_output 匹配）。
+ */
+function appendContinuationMessages(
+  items: ResponsesInputItem[],
+  continuationMessages: ContinuationMessage[],
+): void {
+  // item_id → call_id 映射，供后续 function_call_output 查找
+  const callIdMap = new Map<string, string>()
+
+  for (const contMsg of continuationMessages) {
+    if (contMsg.role === 'assistant') {
+      for (const tc of contMsg.toolCalls) {
+        const callId = tc.metadata?.call_id as string | undefined ?? tc.id
+        callIdMap.set(tc.id, callId)
+        items.push({
+          type: 'function_call',
+          id: tc.id,
+          call_id: callId,
+          name: tc.name,
+          arguments: JSON.stringify(tc.arguments),
+        })
+      }
+    } else if (contMsg.role === 'tool') {
+      for (const result of contMsg.results) {
+        // 优先用 callIdMap 查找真实 call_id，降级直接使用 toolCallId
+        const callId = callIdMap.get(result.toolCallId) ?? result.toolCallId
+        items.push({
+          type: 'function_call_output',
+          call_id: callId,
+          output: result.content,
+        })
+      }
+    }
+  }
+}
+
+// ===== 适配器实现 =====
+
+export class ResponsesAdapter implements ProviderAdapter {
+  readonly providerType = 'openai-responses' as const
+
+  buildStreamRequest(input: StreamRequestInput): ProviderRequest {
+    const url = normalizeBaseUrl(input.baseUrl)
+    const inputItems = toResponsesInput(input)
+
+    // 工具续接消息追加到 input 数组（Responses API 无独立的续接轮次概念）
+    if (input.continuationMessages && input.continuationMessages.length > 0) {
+      appendContinuationMessages(inputItems, input.continuationMessages)
+    }
+
+    const body: Record<string, unknown> = {
+      model: input.modelId,
+      input: inputItems,
+      stream: true,
+    }
+
+    // 系统提示通过 instructions 字段传递（区别于 Chat Completions 的 system role 消息）
+    if (input.systemMessage) {
+      body.instructions = input.systemMessage
+    }
+
+    // 工具定义（扁平格式）
+    if (input.tools && input.tools.length > 0) {
+      body.tools = toResponsesTools(input.tools)
+    }
+
+    return {
+      url: `${url}/responses`,
+      headers: {
+        'Authorization': `Bearer ${input.apiKey}`,
+        'content-type': 'application/json',
+      },
+      body: JSON.stringify(body),
+    }
+  }
+
+  parseSSELine(jsonLine: string): StreamEvent[] {
+    try {
+      const raw = JSON.parse(jsonLine) as ResponsesBaseEvent
+      const events: StreamEvent[] = []
+
+      // 文本增量
+      if (raw.type === 'response.output_text.delta') {
+        const event = raw as ResponsesTextDeltaEvent
+        if (event.delta) {
+          events.push({ type: 'chunk', delta: event.delta })
+        }
+      }
+
+      // 推理文本增量（o1/o3/o4 系列推理模型）
+      else if (raw.type === 'response.reasoning_text.delta') {
+        const event = raw as ResponsesReasoningDeltaEvent
+        if (event.delta) {
+          events.push({ type: 'reasoning', delta: event.delta })
+        }
+      }
+
+      // 新输出项（函数调用开始）
+      else if (raw.type === 'response.output_item.added') {
+        const event = raw as ResponsesOutputItemAddedEvent
+        if (event.item.type === 'function_call') {
+          // item.id 作为跟踪键，与后续 delta 的 item_id 对应
+          const itemId = event.item.id ?? `tc_${event.output_index}`
+          const callId = event.item.call_id ?? itemId
+          events.push({
+            type: 'tool_call_start',
+            toolCallId: itemId,
+            toolName: event.item.name ?? '',
+            // 保留 call_id 供 appendContinuationMessages 构建 function_call_output 使用
+            metadata: { call_id: callId },
+          })
+        }
+      }
+
+      // 函数调用参数增量（item_id 与 tool_call_start 的 toolCallId 对应）
+      else if (raw.type === 'response.function_call_arguments.delta') {
+        const event = raw as ResponsesFunctionCallArgsDeltaEvent
+        if (event.delta) {
+          events.push({
+            type: 'tool_call_delta',
+            toolCallId: event.item_id,
+            argumentsDelta: event.delta,
+          })
+        }
+      }
+
+      return events
+    } catch {
+      return []
+    }
+  }
+
+  buildTitleRequest(input: TitleRequestInput): ProviderRequest {
+    const url = normalizeBaseUrl(input.baseUrl)
+
+    return {
+      url: `${url}/responses`,
+      headers: {
+        'Authorization': `Bearer ${input.apiKey}`,
+        'content-type': 'application/json',
+      },
+      body: JSON.stringify({
+        model: input.modelId,
+        input: [{ role: 'user', content: input.prompt }],
+        stream: false,
+        max_output_tokens: 50,
+      }),
+    }
+  }
+
+  parseTitleResponse(responseBody: unknown): string | null {
+    // 格式：output[0].content[0].text
+    const data = responseBody as { output?: Array<{ content?: Array<{ text?: string }> }> }
+    return data.output?.[0]?.content?.[0]?.text ?? null
+  }
+}
diff --git a/packages/shared/src/types/channel.ts b/packages/shared/src/types/channel.ts
index f5220586..e309220d 100644
--- a/packages/shared/src/types/channel.ts
+++ b/packages/shared/src/types/channel.ts
@@ -11,6 +11,7 @@
 export type ProviderType =
   | 'anthropic'
   | 'openai'
+  | 'openai-responses'
   | 'deepseek'
   | 'google'
   | 'moonshot'
@@ -26,6 +27,7 @@ export type ProviderType =
 export const PROVIDER_DEFAULT_URLS: Record<ProviderType, string> = {
   anthropic: 'https://api.anthropic.com',
   openai: 'https://api.openai.com/v1',
+  'openai-responses': 'https://api.openai.com/v1',
   deepseek: 'https://api.deepseek.com',
   google: 'https://generativelanguage.googleapis.com',
   moonshot: 'https://api.moonshot.cn/v1',
@@ -42,6 +44,7 @@ export const PROVIDER_DEFAULT_URLS: Record<ProviderType, string> = {
 export const PROVIDER_LABELS: Record<ProviderType, string> = {
   anthropic: 'Anthropic',
   openai: 'OpenAI',
+  'openai-responses': 'OpenAI Responses API',
   deepseek: 'DeepSeek',
   google: 'Google',
   moonshot: 'Moonshot / Kimi',