diff --git a/CHANGELOG.md b/CHANGELOG.md
index 9e0e635d..13acb1d0 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,17 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and
 
 [中文版](CHANGELOG.zh.md) · [README](README.md) · [Contributing](CONTRIBUTING.md)
 
+## [2.8.3] - 2026-11-06
+
+### Added
+
+- **Model-based web search**: `bl search web` adds a model search mode and now uses it as the default implementation.
+- **Web search in chat**: `bl text chat` supports built-in web search via the `enable_search` parameter.
+
+### Changed
+
+- **Profile credentials**: Profiles can now declare credentials per search mode (`model` / `mcp`), and the CLI picks the matching profile credential when switching search modes.
+
 ## [2.8.2] - 2026-11-04
 
 ### Internal
diff --git a/CHANGELOG.zh.md b/CHANGELOG.zh.md
index 23d4e16e..d34887a8 100644
--- a/CHANGELOG.zh.md
+++ b/CHANGELOG.zh.md
@@ -6,6 +6,17 @@
 
 [English](CHANGELOG.md) · [README](README.zh.md) · [参与贡献](CONTRIBUTING.zh.md)
 
+## [2.8.3] - 2026-11-06
+
+### 新增
+
+- **模型联网搜索**：`bl search web` 新增模型搜索模式（`--mode model`），并改为默认实现。
+- **对话内置联网搜索**：`bl text chat` 新增 `enable_search` 参数，支持内置联网搜索。
+
+### 变更
+
+- **Profile 凭证**：Profile 可按搜索模式（`model` / `mcp`）声明凭证，切换搜索模式时自动选择对应 Profile 凭证。
+
 ## [2.8.2] - 2026-11-04
 
 ### 内部维护
diff --git a/packages/bailian-kb-dsh/package.json b/packages/bailian-kb-dsh/package.json
index 10d54760..7e1b1734 100644
--- a/packages/bailian-kb-dsh/package.json
+++ b/packages/bailian-kb-dsh/package.json
@@ -1,6 +1,6 @@
 {
   "name": "bailian-kb-dsh",
-  "version": "2.8.2",
+  "version": "2.8.3",
   "description": "Bailian knowledge-base tools for DeepSeek Harness: kb_search and kb_chat over the DashScope RAG API, plus the bl CLI management skill.",
   "keywords": [
     "alibaba-cloud",
diff --git a/packages/bailian-memo-dsh/package.json b/packages/bailian-memo-dsh/package.json
index 2c72f8a5..63e538fa 100644
--- a/packages/bailian-memo-dsh/package.json
+++ b/packages/bailian-memo-dsh/package.json
@@ -1,6 +1,6 @@
 {
   "name": "bailian-memo-dsh",
-  "version": "2.8.2",
+  "version": "2.8.3",
   "description": "Bailian personal memory for DeepSeek Harness: automatic recall, silent curation, and user-controlled remember/forget.",
   "keywords": [
     "alibaba-cloud",
diff --git a/packages/cli/package.json b/packages/cli/package.json
index 6af93671..df104826 100644
--- a/packages/cli/package.json
+++ b/packages/cli/package.json
@@ -1,6 +1,6 @@
 {
   "name": "bailian-cli",
-  "version": "2.8.2",
+  "version": "2.8.3",
   "description": "CLI for Aliyun Model Studio (DashScope) AI Platform.",
   "keywords": [
     "agent",
diff --git a/packages/commands/package.json b/packages/commands/package.json
index 5b405a40..48029c74 100644
--- a/packages/commands/package.json
+++ b/packages/commands/package.json
@@ -1,6 +1,6 @@
 {
   "name": "bailian-cli-commands",
-  "version": "2.8.2",
+  "version": "2.8.3",
   "description": "Command library for bailian-cli products (knowledge, memory, media, …). See https://www.npmjs.com/package/bailian-cli for usage.",
   "homepage": "https://bailian.console.aliyun.com/cli",
   "bugs": {
diff --git a/packages/commands/src/commands/search/web.ts b/packages/commands/src/commands/search/web.ts
index 70083d48..9ec70f60 100644
--- a/packages/commands/src/commands/search/web.ts
+++ b/packages/commands/src/commands/search/web.ts
@@ -2,13 +2,32 @@ import {
   defineCommand,
   formatOutput,
   BailianError,
+  chatPath,
   mcpWebSearchPath,
+  type ChatRequest,
+  type ChatResponse,
   type FlagsDef,
   type McpConnectedClient,
+  type McpToolResult,
 } from "bailian-cli-core";
 
 import { rethrowWithWebSearchActivateHint } from "./web-activate-hint.ts";
 
+/** Fallback model for --mode model, mirroring the text-chat default chain. */
+const DEFAULT_SEARCH_MODEL = "qwen3.8-max";
+
+type WebSearchMode = "mcp" | "model";
+
+/**
+ * Resolve the effective search mode. `--list-tools` is an MCP-only concept and
+ * locks the mode to mcp; otherwise default to model (Chat Completions) unless
+ * `--mode mcp` is given.
+ */
+function resolveSearchMode(flags: { mode?: string; listTools?: boolean }): WebSearchMode {
+  if (flags.listTools) return "mcp";
+  return flags.mode === "mcp" ? "mcp" : "model";
+}
+
 const WEB_SEARCH_FLAGS = {
   query: {
     type: "string",
@@ -30,14 +49,39 @@ const WEB_SEARCH_FLAGS = {
       "zh-CN": "列出可用的 MCP 工具并退出",
     },
   },
+  mode: {
+    type: "string",
+    valueHint: "<mcp|model>",
+    choices: ["mcp", "model"] as const,
+    description: {
+      "en-US":
+        "Search mode: model answers via Chat Completions with enable_search, mcp calls the WebSearch MCP service (default: model)",
+      "zh-CN":
+        "搜索模式：model 通过 Chat Completions（enable_search）由模型联网作答，mcp 调用 WebSearch MCP 服务（默认：model）",
+    },
+  },
+  model: {
+    type: "string",
+    valueHint: "<model>",
+    description: {
+      "en-US": "Model ID for --mode model (default: qwen3.8-max)",
+      "zh-CN": "--mode model 使用的模型 ID（默认：qwen3.8-max）",
+    },
+  },
 } satisfies FlagsDef;
 
 export default defineCommand({
   description: {
-    "en-US": "Search the web using DashScope MCP WebSearch service",
-    "zh-CN": "使用 DashScope MCP WebSearch 服务搜索互联网",
+    "en-US":
+      "Search the web via a chat model with built-in web search (default) or the DashScope MCP WebSearch service",
+    "zh-CN": "搜索互联网：默认走模型内置联网搜索，也可切换为 DashScope MCP WebSearch 服务",
   },
   auth: "apiKey",
+  // Model mode talks Chat Completions (the `text.chat` capability domain), so
+  // Profile API-key selection must follow the mode; MCP mode keeps the
+  // `search.web` capability, which Token Plan profiles deliberately do not
+  // list — plan keys cannot authorize Bailian MCP search.
+  authCapability: (flags) => (resolveSearchMode(flags) === "model" ? "text.chat" : "search.web"),
   usageArgs: "--query <text> [flags]",
   flags: WEB_SEARCH_FLAGS,
   exampleArgs: [
@@ -45,21 +89,79 @@ export default defineCommand({
       "en-US": '--query "Alibaba Cloud Bailian latest features"',
       "zh-CN": '--query "阿里云百炼最新功能"',
     },
-    {
-      "en-US": '--query "TypeScript 5.9 new features" --count 5',
-      "zh-CN": '--query "TypeScript 5.9 新功能" --count 5',
-    },
     {
       "en-US": '--query "Today\'s news"',
       "zh-CN": '--query "今日新闻"',
     },
+    {
+      "en-US": '--mode mcp --query "TypeScript 5.9 new features" --count 5',
+      "zh-CN": '--mode mcp --query "TypeScript 5.9 新功能" --count 5',
+    },
+    {
+      "en-US": '--mode mcp --query "Alibaba Cloud Bailian latest features"',
+      "zh-CN": '--mode mcp --query "阿里云百炼最新功能"',
+    },
     "--list-tools",
   ],
-  validate: (f) => (!f.listTools && !f.query ? "Missing required flag: --query" : undefined),
+  notes: [
+    {
+      "en-US":
+        "Defaults to --mode model: one non-streaming Chat Completions request with enable_search: true, returning the model answer as a single content[].text block (no pages list). Use --mode mcp for the WebSearch MCP service (pages list); --count and --list-tools are MCP-mode only. The model resolves from --model, default_text_model, then qwen3.8-max.",
+      "zh-CN":
+        "默认 --mode model：发送单次非流式 Chat Completions 请求（enable_search: true），模型回答以单个 content[].text 块返回（无 pages 列表）。--mode mcp 走 WebSearch MCP 服务（返回 pages）；--count 与 --list-tools 仅限 MCP 模式。模型按 --model、default_text_model、qwen3.8-max 依次解析。",
+    },
+  ],
+  validate: (flags) => {
+    if (flags.mode === "model" && flags.listTools) {
+      return "--list-tools is only available in MCP mode (--mode mcp). / --list-tools 仅支持 MCP 模式（--mode mcp）。";
+    }
+    if (resolveSearchMode(flags) === "model" && flags.count !== undefined) {
+      return "--count is only available in MCP mode (--mode mcp). / --count 仅支持 MCP 模式（--mode mcp）。";
+    }
+    if (!flags.listTools && !flags.query) return "Missing required flag: --query";
+    return undefined;
+  },
   async run(ctx) {
     const { settings, flags } = ctx;
     const format = settings.output;
 
+    // --- Model mode (default): Chat Completions with enable_search ---
+    if (resolveSearchMode(flags) === "model") {
+      const model = flags.model || settings.defaultTextModel || DEFAULT_SEARCH_MODEL;
+      const request: ChatRequest = {
+        model,
+        messages: [{ role: "user", content: flags.query! }],
+        enable_search: true,
+        stream: false,
+      };
+
+      if (settings.dryRun) {
+        ctx.output.result({ endpoint: ctx.client.url(chatPath()), request });
+        return;
+      }
+
+      ctx.output.progress({ "en-US": "Searching...", "zh-CN": "正在搜索……" });
+      try {
+        const response = await ctx.client.requestJson<ChatResponse>({
+          path: chatPath(),
+          method: "POST",
+          body: request,
+        });
+        const answer = response.choices?.[0]?.message?.content ?? "";
+        ctx.output.progress({ "en-US": "Done.", "zh-CN": "已完成。" });
+        // Keep the MCP result envelope so downstream parsers work for both modes.
+        const result: McpToolResult = {
+          content: [{ type: "text", text: answer }],
+          isError: false,
+        };
+        ctx.output.result(result, format === "text" ? { text: answer } : undefined);
+      } catch (error) {
+        ctx.output.progress({ "en-US": "Failed.", "zh-CN": "失败。" });
+        throw error;
+      }
+      return;
+    }
+
     // --- List tools mode ---
     if (flags.listTools) {
       if (settings.dryRun) {
diff --git a/packages/commands/src/commands/text/chat.ts b/packages/commands/src/commands/text/chat.ts
index d4937c9c..457e6618 100644
--- a/packages/commands/src/commands/text/chat.ts
+++ b/packages/commands/src/commands/text/chat.ts
@@ -25,6 +25,9 @@ import {
 } from "./responses.ts";
 import { PRIME_FLAGS, resolvePrimeEndpoint, validatePrimeFlags } from "../shared/prime.ts";
 
+/** Models that enable built-in web search by default on the Chat Completions API. */
+const SEARCH_BY_DEFAULT_MODELS = new Set(["qwen3.8-max"]);
+
 const CHAT_FLAGS = {
   api: {
     type: "string",
@@ -79,6 +82,15 @@ const CHAT_FLAGS = {
     valueHint: "<n>",
     description: { "en-US": "Nucleus sampling threshold", "zh-CN": "核采样阈值" },
   },
+  enableSearch: {
+    type: "boolean",
+    valueHint: "<true|false>",
+    description: {
+      "en-US":
+        "Enable built-in web search (Chat Completions API only; default: true for qwen3.8-max)",
+      "zh-CN": "启用模型内置联网搜索（仅 Chat Completions API；qwen3.8-max 默认开启）",
+    },
+  },
   stream: {
     type: "switch",
     description: {
@@ -202,6 +214,10 @@ export default defineCommand({
       "en-US": '--model qwq-plus --message "Solve 1+1" --enable-thinking',
       "zh-CN": '--model qwq-plus --message "计算 1+1" --enable-thinking',
     },
+    {
+      "en-US": '--enable-search true --message "Latest news about Alibaba Cloud"',
+      "zh-CN": '--enable-search true --message "阿里云最新新闻"',
+    },
     {
       "en-US": '--prime --workspace-id <id> --model glm-5.3-prime --message "Explain this code"',
       "zh-CN": '--prime --workspace-id <id> --model glm-5.3-prime --message "解释这段代码"',
@@ -229,6 +245,12 @@ export default defineCommand({
       "zh-CN":
         "Prime Endpoint：已配置的 --base-url、DASHSCOPE_BASE_URL 或 Profile base_url 优先；否则按 --workspace-id、BAILIAN_WORKSPACE_ID、配置项 workspace_id 解析工作空间。",
     },
+    {
+      "en-US":
+        '--enable-search maps to the Chat Completions enable_search parameter and requires a model with built-in web search support; whitelisted models (qwen3.8-max) enable it by default unless --enable-search false is passed; for the Responses API use --tool \'{"type":"web_search"}\' instead.',
+      "zh-CN":
+        '--enable-search 映射为 Chat Completions 的 enable_search 参数，需要模型支持内置联网搜索；白名单模型（qwen3.8-max）默认开启，可传 --enable-search false 显式关闭；Responses API 请改用 --tool \'{"type":"web_search"}\'。',
+    },
   ],
   validate: (flags) => {
     if (!flags.message && !flags.messagesFile) return "Provide --message or --messages-file.";
@@ -243,6 +265,9 @@ export default defineCommand({
     if (flags.api === "responses" && flags.thinkingBudget !== undefined) {
       return "--thinking-budget is not supported by the Responses API.";
     }
+    if (flags.api === "responses" && flags.enableSearch !== undefined) {
+      return '--enable-search is not supported by the Responses API; use --tool \'{"type":"web_search"}\' instead. / --enable-search 不支持 Responses API，请改用 --tool \'{"type":"web_search"}\'。';
+    }
     return undefined;
   },
   async run(ctx) {
@@ -295,6 +320,15 @@ export default defineCommand({
       }
     }
 
+    if (api === "chat" && "messages" in body) {
+      if (flags.enableSearch !== undefined) {
+        body.enable_search = flags.enableSearch;
+      } else if (SEARCH_BY_DEFAULT_MODELS.has(model)) {
+        // Whitelisted models get built-in web search unless explicitly opted out.
+        body.enable_search = true;
+      }
+    }
+
     const requestPath = flags.batch
       ? resolveBatchChatEndpoint(ctx.client)
       : flags.prime
diff --git a/packages/commands/tests/e2e/search-web.e2e.test.ts b/packages/commands/tests/e2e/search-web.e2e.test.ts
index c924e10e..c5c9ba83 100644
--- a/packages/commands/tests/e2e/search-web.e2e.test.ts
+++ b/packages/commands/tests/e2e/search-web.e2e.test.ts
@@ -1,5 +1,13 @@
+import { writeFileSync } from "node:fs";
+import { join } from "node:path";
 import { describe, expect, test } from "vite-plus/test";
-import { isDashScopeE2EReady, parseStdoutJson, runCommandHelp, runCommandE2e } from "./helpers.ts";
+import {
+  isDashScopeE2EReady,
+  makeE2eOutputDir,
+  parseStdoutJson,
+  runCommandHelp,
+  runCommandE2e,
+} from "./helpers.ts";
 import { SEARCH_WEB_ROUTES } from "./topic-routes.ts";
 
 function pagesFromSearchWebStdout(stdout: string): Array<{ title?: string; url?: string }> {
@@ -23,7 +31,7 @@ describe("e2e: search web", () => {
       "--help",
     ]);
     expect(exitCode, stderr).toBe(0);
-    expect(stderr).toMatch(/web|--query|list-tools|count/i);
+    expect(stderr).toMatch(/web|--query|list-tools|count|--mode/i);
   });
 
   test("search web --dry-run --list-tools 无需 --query 也无需凭证即可干跑", async () => {
@@ -39,6 +47,177 @@ describe("e2e: search web", () => {
     const data = parseStdoutJson<{ action?: string }>(stdout);
     expect(data.action).toBe("tools/list");
   });
+
+  test("search web --mode model --dry-run 生成 enable_search 对话请求", async () => {
+    const { stdout, stderr, exitCode } = await runCommandE2e(
+      SEARCH_WEB_ROUTES,
+      ["search", "web", "--mode", "model", "--dry-run", "--output", "json", "--query", "干跑校验"],
+      {
+        DASHSCOPE_API_KEY: undefined,
+        DASHSCOPE_ACCESS_TOKEN: undefined,
+      },
+    );
+    expect(exitCode, stderr).toBe(0);
+    const data = parseStdoutJson<{
+      endpoint?: string;
+      request?: {
+        model?: string;
+        messages?: Array<{ role?: string; content?: string }>;
+        enable_search?: boolean;
+        stream?: boolean;
+      };
+    }>(stdout);
+    expect(data.endpoint).toMatch(/\/chat\/completions$/);
+    expect(data.request?.model).toBe("qwen3.8-max");
+    expect(data.request?.messages).toEqual([{ role: "user", content: "干跑校验" }]);
+    expect(data.request?.enable_search).toBe(true);
+    expect(data.request?.stream).toBe(false);
+  });
+
+  test("search web --mode model 可用 --model 指定模型", async () => {
+    const { stdout, stderr, exitCode } = await runCommandE2e(
+      SEARCH_WEB_ROUTES,
+      [
+        "search",
+        "web",
+        "--mode",
+        "model",
+        "--model",
+        "qwen-max",
+        "--dry-run",
+        "--output",
+        "json",
+        "--query",
+        "干跑校验",
+      ],
+      {
+        DASHSCOPE_API_KEY: undefined,
+        DASHSCOPE_ACCESS_TOKEN: undefined,
+      },
+    );
+    expect(exitCode, stderr).toBe(0);
+    const data = parseStdoutJson<{ request?: { model?: string } }>(stdout);
+    expect(data.request?.model).toBe("qwen-max");
+  });
+
+  test("search web --mode model 拒绝 --list-tools 与 --count", async () => {
+    const listTools = await runCommandE2e(SEARCH_WEB_ROUTES, [
+      "search",
+      "web",
+      "--mode",
+      "model",
+      "--list-tools",
+    ]);
+    expect(listTools.exitCode).toBe(2);
+    expect(listTools.stderr).toMatch(/--list-tools.*MCP/i);
+
+    const count = await runCommandE2e(SEARCH_WEB_ROUTES, [
+      "search",
+      "web",
+      "--mode",
+      "model",
+      "--query",
+      "hello",
+      "--count",
+      "3",
+    ]);
+    expect(count.exitCode).toBe(2);
+    expect(count.stderr).toMatch(/--count.*MCP/i);
+  });
+
+  test("search web 拒绝未知的 --mode 值", async () => {
+    const { stderr, exitCode } = await runCommandE2e(SEARCH_WEB_ROUTES, [
+      "search",
+      "web",
+      "--mode",
+      "legacy",
+      "--query",
+      "hello",
+    ]);
+    expect(exitCode).toBe(2);
+    expect(stderr).toMatch(/--mode|mcp.*model/i);
+  });
+
+  test("search web 默认走 model：--dry-run 生成 enable_search 对话请求", async () => {
+    const { stdout, stderr, exitCode } = await runCommandE2e(
+      SEARCH_WEB_ROUTES,
+      ["search", "web", "--dry-run", "--output", "json", "--query", "干跑校验"],
+      {
+        DASHSCOPE_API_KEY: undefined,
+        DASHSCOPE_ACCESS_TOKEN: undefined,
+      },
+    );
+    expect(exitCode, stderr).toBe(0);
+    const data = parseStdoutJson<{
+      endpoint?: string;
+      request?: { model?: string; enable_search?: boolean };
+    }>(stdout);
+    expect(data.endpoint).toMatch(/\/chat\/completions$/);
+    expect(data.request?.model).toBe("qwen3.8-max");
+    expect(data.request?.enable_search).toBe(true);
+  });
+
+  test("search web 默认 model 模式拒绝 --count", async () => {
+    const { stderr, exitCode } = await runCommandE2e(SEARCH_WEB_ROUTES, [
+      "search",
+      "web",
+      "--query",
+      "hello",
+      "--count",
+      "3",
+    ]);
+    expect(exitCode).toBe(2);
+    expect(stderr).toMatch(/--count.*MCP/i);
+  });
+
+  test("search web 按模式选择 Profile 凭证:model 保留 token-plan,mcp 回退 default", async () => {
+    const configDir = makeE2eOutputDir("search-web-mode-capability");
+    writeFileSync(
+      join(configDir, "config.json"),
+      JSON.stringify(
+        {
+          api_key: "sk-default",
+          base_url: "https://default.example.com",
+          active_config: "token-plan",
+          "token-plan": {
+            api_key: "sk-token-plan",
+            base_url: "https://token-plan.example.com",
+            api_key_capabilities: ["text.chat"],
+          },
+        },
+        null,
+        2,
+      ) + "\n",
+    );
+    const env = {
+      BAILIAN_CONFIG_DIR: configDir,
+      DASHSCOPE_API_KEY: undefined,
+      DASHSCOPE_BASE_URL: undefined,
+      DASHSCOPE_ACCESS_TOKEN: undefined,
+    };
+
+    // Model mode authenticates as `text.chat`, which the token-plan Profile
+    // lists — its connection must survive capability selection.
+    const model = await runCommandE2e(
+      SEARCH_WEB_ROUTES,
+      ["search", "web", "--dry-run", "--output", "json", "--query", "凭证选择"],
+      env,
+    );
+    expect(model.exitCode, model.stderr).toBe(0);
+    const modelData = parseStdoutJson<{ endpoint?: string }>(model.stdout);
+    expect(modelData.endpoint).toContain("https://token-plan.example.com");
+
+    // MCP mode keeps the `search.web` capability, absent from token-plan —
+    // the run must fall back to the default Profile connection.
+    const mcp = await runCommandE2e(
+      SEARCH_WEB_ROUTES,
+      ["search", "web", "--mode", "mcp", "--dry-run", "--output", "json", "--query", "凭证选择"],
+      env,
+    );
+    expect(mcp.exitCode, mcp.stderr).toBe(0);
+    const mcpData = parseStdoutJson<{ endpoint?: string }>(mcp.stdout);
+    expect(mcpData.endpoint).toContain("https://default.example.com");
+  });
 });
 
 describe.skipIf(!isDashScopeE2EReady())("e2e: search web", () => {
@@ -52,10 +231,12 @@ describe.skipIf(!isDashScopeE2EReady())("e2e: search web", () => {
     expect(stderr).toMatch(/--query|Usage:/i);
   });
 
-  test("search web --dry-run 仅输出计划且不调 MCP", async () => {
+  test("search web --mode mcp --dry-run 仅输出 tools/call 计划", async () => {
     const { stdout, stderr, exitCode } = await runCommandE2e(SEARCH_WEB_ROUTES, [
       "search",
       "web",
+      "--mode",
+      "mcp",
       "--dry-run",
       "--output",
       "json",
@@ -76,10 +257,12 @@ describe.skipIf(!isDashScopeE2EReady())("e2e: search web", () => {
     expect(data.arguments?.count).toBe(5);
   });
 
-  test("联网搜索返回 JSON 且含搜索结果", async () => {
+  test("MCP 联网搜索返回 JSON 且含 pages", async () => {
     const { stdout, stderr, exitCode } = await runCommandE2e(SEARCH_WEB_ROUTES, [
       "search",
       "web",
+      "--mode",
+      "mcp",
       "--query",
       "阿里云百炼",
       "--count",
@@ -107,4 +290,23 @@ describe.skipIf(!isDashScopeE2EReady())("e2e: search web", () => {
     const data = parseStdoutJson<{ tools?: Array<{ name?: string }> }>(stdout);
     expect(data.tools?.some((tool) => tool.name === "bailian_web_search")).toBe(true);
   }, 60_000);
+
+  test("--mode model 真实调用模型联网搜索并返回 content 信封", async () => {
+    const { stdout, stderr, exitCode } = await runCommandE2e(SEARCH_WEB_ROUTES, [
+      "search",
+      "web",
+      "--mode",
+      "model",
+      "--query",
+      "阿里云百炼是什么？一句话回答",
+      "--output",
+      "json",
+    ]);
+    expect(exitCode, stderr).toBe(0);
+    // 模型模式无 pages；content[].text 直接承载模型回答纯文本（非 JSON）。
+    const envelope = parseStdoutJson<{ content?: Array<{ type?: string; text?: string }> }>(stdout);
+    const text = envelope.content?.find((block) => block.type === "text")?.text ?? "";
+    expect(text.length).toBeGreaterThan(0);
+    expect(() => JSON.parse(text)).toThrow();
+  }, 120_000);
 });
diff --git a/packages/commands/tests/e2e/text-chat.e2e.test.ts b/packages/commands/tests/e2e/text-chat.e2e.test.ts
index 1fbf95dc..f9050413 100644
--- a/packages/commands/tests/e2e/text-chat.e2e.test.ts
+++ b/packages/commands/tests/e2e/text-chat.e2e.test.ts
@@ -2,7 +2,13 @@ import { mkdtempSync, rmSync, writeFileSync } from "node:fs";
 import { tmpdir } from "node:os";
 import { join } from "node:path";
 import { afterEach, beforeEach, describe, expect, test } from "vite-plus/test";
-import { isDashScopeE2EReady, parseStdoutJson, runCommandHelp, runCommandE2e } from "./helpers.ts";
+import {
+  isDashScopeE2EReady,
+  parseStdoutJson,
+  runCommandHelp,
+  runCommandE2e,
+  runCommandInProcess,
+} from "./helpers.ts";
 import { TEXT_CHAT_ROUTES } from "./topic-routes.ts";
 
 let isolatedDirectory: string;
@@ -40,6 +46,7 @@ describe("e2e: text chat", () => {
     const { stderr, exitCode } = await runCommandHelp(TEXT_CHAT_ROUTES, ["text", "chat", "--help"]);
     expect(exitCode, stderr).toBe(0);
     expect(stderr).toMatch(/--api\s+<chat\|responses>/i);
+    expect(stderr).toMatch(/--enable-search/i);
     expect(stderr).toMatch(/--prime/i);
     expect(stderr).toMatch(/--workspace-id/i);
     expect(stderr).toMatch(/chat|--message|model|stream/i);
@@ -73,6 +80,140 @@ describe("e2e: text chat", () => {
     expect(stderr).toMatch(/thinking-budget.*Responses/i);
   });
 
+  test("text chat 在 Responses 模式拒绝 --enable-search", async () => {
+    const { stderr, exitCode } = await runCommandInProcess(TEXT_CHAT_ROUTES, [
+      "text",
+      "chat",
+      "--api",
+      "responses",
+      "--message",
+      "hello",
+      "--enable-search",
+      "true",
+    ]);
+    expect(exitCode).toBe(2);
+    expect(stderr).toMatch(/enable-search.*Responses/i);
+  });
+
+  test("text chat --dry-run 透传 --enable-search true/false 到请求体", async () => {
+    const { stdout, stderr, exitCode } = await runCommandInProcess(
+      TEXT_CHAT_ROUTES,
+      [
+        "text",
+        "chat",
+        "--dry-run",
+        "--enable-search",
+        "true",
+        "--message",
+        "hello",
+        "--output",
+        "json",
+      ],
+      { DASHSCOPE_API_KEY: "" },
+    );
+    expect(exitCode, stderr).toBe(0);
+    const enabled = parseStdoutJson<{ request?: { enable_search?: boolean } }>(stdout);
+    expect(enabled.request?.enable_search).toBe(true);
+
+    const {
+      stdout: disabledStdout,
+      stderr: disabledStderr,
+      exitCode: disabledExit,
+    } = await runCommandInProcess(
+      TEXT_CHAT_ROUTES,
+      [
+        "text",
+        "chat",
+        "--dry-run",
+        "--enable-search",
+        "false",
+        "--message",
+        "hello",
+        "--output",
+        "json",
+      ],
+      { DASHSCOPE_API_KEY: "" },
+    );
+    expect(disabledExit, disabledStderr).toBe(0);
+    const disabled = parseStdoutJson<{ request?: { enable_search?: boolean } }>(disabledStdout);
+    expect(disabled.request?.enable_search).toBe(false);
+  });
+
+  test("text chat --dry-run 未传 --enable-search 时请求体不含 enable_search", async () => {
+    const { stdout, stderr, exitCode } = await runCommandInProcess(
+      TEXT_CHAT_ROUTES,
+      [
+        "text",
+        "chat",
+        "--dry-run",
+        "--model",
+        "qwen-max",
+        "--message",
+        "hello",
+        "--output",
+        "json",
+      ],
+      { DASHSCOPE_API_KEY: "" },
+    );
+    expect(exitCode, stderr).toBe(0);
+    const data = parseStdoutJson<{ request?: Record<string, unknown> }>(stdout);
+    expect(data.request).not.toHaveProperty("enable_search");
+  });
+
+  test("text chat --dry-run 白名单模型 qwen3.8-max 默认开启 enable_search", async () => {
+    // 未显式传 --model 时兼容默认模型也是 qwen3.8-max，同样命中白名单。
+    const { stdout, stderr, exitCode } = await runCommandInProcess(
+      TEXT_CHAT_ROUTES,
+      [
+        "text",
+        "chat",
+        "--dry-run",
+        "--model",
+        "qwen3.8-max",
+        "--message",
+        "hello",
+        "--output",
+        "json",
+      ],
+      { DASHSCOPE_API_KEY: "" },
+    );
+    expect(exitCode, stderr).toBe(0);
+    const data = parseStdoutJson<{ request?: { enable_search?: boolean } }>(stdout);
+    expect(data.request?.enable_search).toBe(true);
+
+    const { stdout: implicitStdout, exitCode: implicitExit } = await runCommandInProcess(
+      TEXT_CHAT_ROUTES,
+      ["text", "chat", "--dry-run", "--message", "hello", "--output", "json"],
+      { DASHSCOPE_API_KEY: "" },
+    );
+    expect(implicitExit).toBe(0);
+    const implicit = parseStdoutJson<{ request?: { enable_search?: boolean } }>(implicitStdout);
+    expect(implicit.request?.enable_search).toBe(true);
+  });
+
+  test("text chat --dry-run 白名单模型可用 --enable-search false 显式关闭", async () => {
+    const { stdout, stderr, exitCode } = await runCommandInProcess(
+      TEXT_CHAT_ROUTES,
+      [
+        "text",
+        "chat",
+        "--dry-run",
+        "--model",
+        "qwen3.8-max",
+        "--enable-search",
+        "false",
+        "--message",
+        "hello",
+        "--output",
+        "json",
+      ],
+      { DASHSCOPE_API_KEY: "" },
+    );
+    expect(exitCode, stderr).toBe(0);
+    const data = parseStdoutJson<{ request?: { enable_search?: boolean } }>(stdout);
+    expect(data.request?.enable_search).toBe(false);
+  });
+
   test("Prime 模式要求显式 --model", async () => {
     const { stderr, exitCode } = await runIsolatedTextChat(
       ["text", "chat", "--prime", "--message", "hello"],
diff --git a/packages/commands/tests/output/fixtures/inference-output.json b/packages/commands/tests/output/fixtures/inference-output.json
index ffbfe234..ced89ba8 100644
--- a/packages/commands/tests/output/fixtures/inference-output.json
+++ b/packages/commands/tests/output/fixtures/inference-output.json
@@ -1669,221 +1669,99 @@
     ]
   },
   "search:text:false": {
-    "stdout": "pages:\n  - title: Fixture\n    url: https://example.com/fixture\n    snippet: fixture snippet\n    hostname: example.com\ntotal: 1\ntext: extra result\n",
+    "stdout": "Hello fixture\n",
     "requests": [
       {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
-        "method": "POST",
-        "body": {
-          "jsonrpc": "2.0",
-          "id": 0,
-          "method": "initialize",
-          "params": {
-            "protocolVersion": "2025-11-25",
-            "capabilities": {},
-            "clientInfo": {
-              "name": "test",
-              "version": "0.0.0"
-            }
-          }
-        }
-      },
-      {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
-        "method": "POST",
-        "body": {
-          "jsonrpc": "2.0",
-          "method": "notifications/initialized"
-        }
-      },
-      {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
+        "url": "https://fixture.invalid/compatible-mode/v1/chat/completions",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "id": 1,
-          "method": "tools/call",
-          "params": {
-            "name": "bailian_web_search",
-            "arguments": {
-              "query": "fixture query",
-              "count": 2
+          "model": "qwen3.8-max",
+          "messages": [
+            {
+              "role": "user",
+              "content": "fixture query"
             }
-          }
+          ],
+          "enable_search": true,
+          "stream": false
         }
       }
     ]
   },
   "search:json:false": {
-    "stdout": "{\n  \"isError\": false,\n  \"content\": [\n    {\n      \"type\": \"text\",\n      \"text\": \"{\\\"pages\\\":[{\\\"title\\\":\\\"Fixture\\\",\\\"url\\\":\\\"https://example.com/fixture\\\",\\\"snippet\\\":\\\"fixture snippet\\\",\\\"hostname\\\":\\\"example.com\\\"}]}\"\n    },\n    {\n      \"type\": \"text\",\n      \"text\": \"extra result\"\n    }\n  ],\n  \"structuredContent\": {\n    \"future\": {\n      \"values\": [\n        false,\n        0,\n        null,\n        []\n      ]\n    }\n  }\n}\n",
+    "stdout": "{\n  \"content\": [\n    {\n      \"type\": \"text\",\n      \"text\": \"Hello fixture\"\n    }\n  ],\n  \"isError\": false\n}\n",
     "data": {
-      "isError": false,
       "content": [
         {
           "type": "text",
-          "text": "{\"pages\":[{\"title\":\"Fixture\",\"url\":\"https://example.com/fixture\",\"snippet\":\"fixture snippet\",\"hostname\":\"example.com\"}]}"
-        },
-        {
-          "type": "text",
-          "text": "extra result"
+          "text": "Hello fixture"
         }
       ],
-      "structuredContent": {
-        "future": {
-          "values": [false, 0, null, []]
-        }
-      }
+      "isError": false
     },
     "requests": [
       {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
-        "method": "POST",
-        "body": {
-          "jsonrpc": "2.0",
-          "id": 0,
-          "method": "initialize",
-          "params": {
-            "protocolVersion": "2025-11-25",
-            "capabilities": {},
-            "clientInfo": {
-              "name": "test",
-              "version": "0.0.0"
-            }
-          }
-        }
-      },
-      {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
-        "method": "POST",
-        "body": {
-          "jsonrpc": "2.0",
-          "method": "notifications/initialized"
-        }
-      },
-      {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
+        "url": "https://fixture.invalid/compatible-mode/v1/chat/completions",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "id": 1,
-          "method": "tools/call",
-          "params": {
-            "name": "bailian_web_search",
-            "arguments": {
-              "query": "fixture query",
-              "count": 2
+          "model": "qwen3.8-max",
+          "messages": [
+            {
+              "role": "user",
+              "content": "fixture query"
             }
-          }
+          ],
+          "enable_search": true,
+          "stream": false
         }
       }
     ]
   },
   "search:text:true": {
-    "stdout": "pages:\n  - title: Fixture\n    url: https://example.com/fixture\n    snippet: fixture snippet\n    hostname: example.com\ntotal: 1\ntext: extra result\n",
+    "stdout": "Hello fixture\n",
     "requests": [
       {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
-        "method": "POST",
-        "body": {
-          "jsonrpc": "2.0",
-          "id": 0,
-          "method": "initialize",
-          "params": {
-            "protocolVersion": "2025-11-25",
-            "capabilities": {},
-            "clientInfo": {
-              "name": "test",
-              "version": "0.0.0"
-            }
-          }
-        }
-      },
-      {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
-        "method": "POST",
-        "body": {
-          "jsonrpc": "2.0",
-          "method": "notifications/initialized"
-        }
-      },
-      {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
+        "url": "https://fixture.invalid/compatible-mode/v1/chat/completions",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "id": 1,
-          "method": "tools/call",
-          "params": {
-            "name": "bailian_web_search",
-            "arguments": {
-              "query": "fixture query",
-              "count": 2
+          "model": "qwen3.8-max",
+          "messages": [
+            {
+              "role": "user",
+              "content": "fixture query"
             }
-          }
+          ],
+          "enable_search": true,
+          "stream": false
         }
       }
     ]
   },
   "search:json:true": {
-    "stdout": "{\n  \"isError\": false,\n  \"content\": [\n    {\n      \"type\": \"text\",\n      \"text\": \"{\\\"pages\\\":[{\\\"title\\\":\\\"Fixture\\\",\\\"url\\\":\\\"https://example.com/fixture\\\",\\\"snippet\\\":\\\"fixture snippet\\\",\\\"hostname\\\":\\\"example.com\\\"}]}\"\n    },\n    {\n      \"type\": \"text\",\n      \"text\": \"extra result\"\n    }\n  ],\n  \"structuredContent\": {\n    \"future\": {\n      \"values\": [\n        false,\n        0,\n        null,\n        []\n      ]\n    }\n  }\n}\n",
+    "stdout": "{\n  \"content\": [\n    {\n      \"type\": \"text\",\n      \"text\": \"Hello fixture\"\n    }\n  ],\n  \"isError\": false\n}\n",
     "data": {
-      "isError": false,
       "content": [
         {
           "type": "text",
-          "text": "{\"pages\":[{\"title\":\"Fixture\",\"url\":\"https://example.com/fixture\",\"snippet\":\"fixture snippet\",\"hostname\":\"example.com\"}]}"
-        },
-        {
-          "type": "text",
-          "text": "extra result"
+          "text": "Hello fixture"
         }
       ],
-      "structuredContent": {
-        "future": {
-          "values": [false, 0, null, []]
-        }
-      }
+      "isError": false
     },
     "requests": [
       {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
-        "method": "POST",
-        "body": {
-          "jsonrpc": "2.0",
-          "id": 0,
-          "method": "initialize",
-          "params": {
-            "protocolVersion": "2025-11-25",
-            "capabilities": {},
-            "clientInfo": {
-              "name": "test",
-              "version": "0.0.0"
-            }
-          }
-        }
-      },
-      {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
-        "method": "POST",
-        "body": {
-          "jsonrpc": "2.0",
-          "method": "notifications/initialized"
-        }
-      },
-      {
-        "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
+        "url": "https://fixture.invalid/compatible-mode/v1/chat/completions",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "id": 1,
-          "method": "tools/call",
-          "params": {
-            "name": "bailian_web_search",
-            "arguments": {
-              "query": "fixture query",
-              "count": 2
+          "model": "qwen3.8-max",
+          "messages": [
+            {
+              "role": "user",
+              "content": "fixture query"
             }
-          }
+          ],
+          "enable_search": true,
+          "stream": false
         }
       }
     ]
@@ -1895,8 +1773,6 @@
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "id": 0,
           "method": "initialize",
           "params": {
             "protocolVersion": "2025-11-25",
@@ -1905,24 +1781,26 @@
               "name": "test",
               "version": "0.0.0"
             }
-          }
+          },
+          "jsonrpc": "2.0",
+          "id": 0
         }
       },
       {
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "method": "notifications/initialized"
+          "method": "notifications/initialized",
+          "jsonrpc": "2.0"
         }
       },
       {
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
+          "method": "tools/list",
           "jsonrpc": "2.0",
-          "id": 1,
-          "method": "tools/list"
+          "id": 1
         }
       }
     ]
@@ -1945,8 +1823,6 @@
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "id": 0,
           "method": "initialize",
           "params": {
             "protocolVersion": "2025-11-25",
@@ -1955,24 +1831,26 @@
               "name": "test",
               "version": "0.0.0"
             }
-          }
+          },
+          "jsonrpc": "2.0",
+          "id": 0
         }
       },
       {
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "method": "notifications/initialized"
+          "method": "notifications/initialized",
+          "jsonrpc": "2.0"
         }
       },
       {
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
+          "method": "tools/list",
           "jsonrpc": "2.0",
-          "id": 1,
-          "method": "tools/list"
+          "id": 1
         }
       }
     ]
@@ -1984,8 +1862,6 @@
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "id": 0,
           "method": "initialize",
           "params": {
             "protocolVersion": "2025-11-25",
@@ -1994,24 +1870,26 @@
               "name": "test",
               "version": "0.0.0"
             }
-          }
+          },
+          "jsonrpc": "2.0",
+          "id": 0
         }
       },
       {
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "method": "notifications/initialized"
+          "method": "notifications/initialized",
+          "jsonrpc": "2.0"
         }
       },
       {
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
+          "method": "tools/list",
           "jsonrpc": "2.0",
-          "id": 1,
-          "method": "tools/list"
+          "id": 1
         }
       }
     ]
@@ -2034,8 +1912,6 @@
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "id": 0,
           "method": "initialize",
           "params": {
             "protocolVersion": "2025-11-25",
@@ -2044,24 +1920,26 @@
               "name": "test",
               "version": "0.0.0"
             }
-          }
+          },
+          "jsonrpc": "2.0",
+          "id": 0
         }
       },
       {
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
-          "jsonrpc": "2.0",
-          "method": "notifications/initialized"
+          "method": "notifications/initialized",
+          "jsonrpc": "2.0"
         }
       },
       {
         "url": "https://fixture.invalid/api/v1/mcps/WebSearch/mcp",
         "method": "POST",
         "body": {
+          "method": "tools/list",
           "jsonrpc": "2.0",
-          "id": 1,
-          "method": "tools/list"
+          "id": 1
         }
       }
     ]
diff --git a/packages/commands/tests/output/inference-output.test.ts b/packages/commands/tests/output/inference-output.test.ts
index d47ce48d..27e14947 100644
--- a/packages/commands/tests/output/inference-output.test.ts
+++ b/packages/commands/tests/output/inference-output.test.ts
@@ -304,6 +304,7 @@ async function invoke(
       return new Response("", { status: 200 });
     }
     if (scenario.kind === "search") {
+      if (address.includes("/chat/completions")) return Response.json(chat);
       if (String(json.method).startsWith("notifications/"))
         return new Response(null, { status: 204 });
       return Response.json({
diff --git a/packages/core/package.json b/packages/core/package.json
index beb5c49b..fe12a150 100644
--- a/packages/core/package.json
+++ b/packages/core/package.json
@@ -1,6 +1,6 @@
 {
   "name": "bailian-cli-core",
-  "version": "2.8.2",
+  "version": "2.8.3",
   "description": "Core SDK for bailian-cli. See https://www.npmjs.com/package/bailian-cli for usage.",
   "homepage": "https://bailian.console.aliyun.com/cli",
   "bugs": {
diff --git a/packages/core/src/types/api.ts b/packages/core/src/types/api.ts
index b324fb92..2c7e94c1 100644
--- a/packages/core/src/types/api.ts
+++ b/packages/core/src/types/api.ts
@@ -42,6 +42,7 @@ export interface ChatRequest {
   tool_choice?: "auto" | "none" | { type: "function"; function: { name: string } };
   enable_thinking?: boolean;
   thinking_budget?: number;
+  enable_search?: boolean;
   modalities?: string[];
   audio?: { voice: string; format?: string };
   stream_options?: { include_usage?: boolean };
diff --git a/packages/core/src/types/command.ts b/packages/core/src/types/command.ts
index e5e6c131..4844c1e7 100644
--- a/packages/core/src/types/command.ts
+++ b/packages/core/src/types/command.ts
@@ -318,6 +318,15 @@ export interface Command<F extends FlagsDef = FlagsDef, Prepared = undefined> {
   description: LocalizedText;
   /** Credential this command requires. See {@link AuthRequirement}. */
   auth: AuthRequirement;
+  /**
+   * Capability key used for Profile API-key selection (see
+   * `selectApiKeyResolutionSources`). Defaults to the command path joined with
+   * ".". Commands whose network target switches with a flag supply a resolver
+   * so credential selection follows the actual connection: `search web` model
+   * mode talks Chat Completions and authenticates as `text.chat`, while
+   * `--mode mcp` keeps the MCP-only `search.web` capability.
+   */
+  authCapability?: (flags: ParsedFlags<F>) => string;
   /**
    * Best-effort secondary console credential for `auth: "apiKey"` commands.
    * When true, the runtime additionally resolves the console credential and
diff --git a/packages/kscli/package.json b/packages/kscli/package.json
index 799612b5..2d47e2d9 100644
--- a/packages/kscli/package.json
+++ b/packages/kscli/package.json
@@ -1,6 +1,6 @@
 {
   "name": "knowledge-studio-cli",
-  "version": "2.8.2",
+  "version": "2.8.3",
   "description": "Lightweight RAG CLI for Aliyun Model Studio — focused on knowledge-base retrieval.",
   "keywords": [
     "alibaba-cloud",
diff --git a/packages/runtime/package.json b/packages/runtime/package.json
index f137b046..d0ef59ee 100644
--- a/packages/runtime/package.json
+++ b/packages/runtime/package.json
@@ -1,6 +1,6 @@
 {
   "name": "bailian-cli-runtime",
-  "version": "2.8.2",
+  "version": "2.8.3",
   "description": "Runtime framework for bailian-cli (createCli, registry, args, output, pipeline). See https://www.npmjs.com/package/bailian-cli for usage.",
   "homepage": "https://bailian.console.aliyun.com/cli",
   "bugs": {
diff --git a/packages/runtime/src/middleware.ts b/packages/runtime/src/middleware.ts
index 0687a49b..22b8f3ae 100644
--- a/packages/runtime/src/middleware.ts
+++ b/packages/runtime/src/middleware.ts
@@ -195,7 +195,11 @@ export const authStage: Middleware = async (ctx, next) => {
     ...resolveModelBaseUrlState(sources),
   };
   if (command.auth === "apiKey") {
-    const capability = ctx.path.join(".");
+    // Commands whose network target switches with a flag declare authCapability
+    // so Profile selection follows the actual connection; others key off path.
+    const capability = command.authCapability
+      ? command.authCapability(ctx.flags)
+      : ctx.path.join(".");
     const selection = selectApiKeyResolutionSources(sources, capability);
     const apiSources = selection.sources;
     if (selection.fallbackFrom) {
diff --git a/packages/runtime/tests/auth-profile-fallback.test.ts b/packages/runtime/tests/auth-profile-fallback.test.ts
index dc0b7774..de54b88a 100644
--- a/packages/runtime/tests/auth-profile-fallback.test.ts
+++ b/packages/runtime/tests/auth-profile-fallback.test.ts
@@ -12,6 +12,7 @@ import {
   makeConfigStore,
   resolveModelBaseUrl,
   writeConfigFile,
+  type AnyCommand,
   type CommandPackManager,
   type LocalizedText,
   type SourceFlags,
@@ -39,15 +40,15 @@ function makeContext(
   path: string[],
   flags: Partial<SourceFlags> = { config: "company-plan" },
   env: NodeJS.ProcessEnv = {},
+  command: AnyCommand = defineCommand({
+    description: "test command",
+    auth: "apiKey",
+    async run() {},
+  }),
 ): RunContext {
   const sources = buildSources(flags);
   sources.env = env;
   const settings = { ...buildSettings(sources), quiet: true };
-  const command = defineCommand({
-    description: "test command",
-    auth: "apiKey",
-    async run() {},
-  });
   return {
     identity: {
       binName: "bl",
@@ -189,6 +190,43 @@ test("authStage 不为缺少 capability 字段的旧 token-plan 注入内置 pre
   });
 });
 
+test("authStage 按命令 authCapability 选 capability:model 模式保留 token-plan,mcp 模式回退 default", async () => {
+  useTempConfigDir();
+  await writeConfigFile({ api_key: "sk-default", base_url: "https://default.example.com" });
+  await writeConfigFile(
+    {
+      api_key: "sk-token-plan",
+      base_url: "https://token-plan.example.com",
+      api_key_capabilities: ["text.chat"],
+    },
+    "token-plan",
+  );
+  const command = defineCommand({
+    description: "test search",
+    auth: "apiKey",
+    flags: {
+      mode: { type: "string", valueHint: "<mcp|model>", description: "search mode" },
+    },
+    authCapability: (flags) => (flags.mode === "mcp" ? "search.web" : "text.chat"),
+    async run() {},
+  });
+
+  const modelContext = makeContext(["search", "web"], { config: "token-plan" }, {}, command);
+  await runAuth(modelContext);
+  expect(modelContext.client.exportApiCredential()).toMatchObject({
+    token: "sk-token-plan",
+    baseUrl: "https://token-plan.example.com",
+  });
+
+  const mcpContext = makeContext(["search", "web"], { config: "token-plan" }, {}, command);
+  mcpContext.flags = { mode: "mcp" };
+  await runAuth(mcpContext);
+  expect(mcpContext.client.exportApiCredential()).toMatchObject({
+    token: "sk-default",
+    baseUrl: "https://default.example.com",
+  });
+});
+
 test("authStage 在 --api-key 覆盖 Profile API Key 时跳过 capability fallback", async () => {
   useTempConfigDir();
   await writeConfigFile({ api_key: "sk-default", base_url: "https://default.example.com" });
diff --git a/skills/bailian-cli/SKILL.md b/skills/bailian-cli/SKILL.md
index 350f101b..22c4b3ac 100644
--- a/skills/bailian-cli/SKILL.md
+++ b/skills/bailian-cli/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: bailian-cli
 metadata:
-  version: "2.8.2"
+  version: "2.8.3"
   requires:
     bins: ["bl"]
 description: >-
@@ -14,7 +14,7 @@ description: >-
   家族路由：生图/生视频/配音/语音合成/转写 → bailian-gen；精调/微调/训练/训练数据集 → bailian-finetune；
   agents.yaml 托管 Agent → bailian-managed-agent；Sandbox 实例与模版 → bailian-sandbox；
   个人长期记忆与应用记忆资源 → bailian-memory；个人历史资料辅助任务、资料接入与知识库管理 → bailian-knowledge；
-  联网搜索的模型路由（Token Plan 自带搜索 vs MCP 搜索 + 兜底）→ bailian-web-search。
+  联网搜索路由（默认模型搜索 vs 显式 `--mode mcp` + 兜底）→ bailian-web-search。
   不要用于普通问答、编程、写作、翻译、摘要、泛搜索，或图片理解等宿主自己能做的任务（普通问答、编程、写作、翻译、摘要、泛搜索不触发）。
   未命名用量/额度问题：先问用户使用哪个产品，再运行 `bl usage` / `bl quota` 查询。
 ---
@@ -101,7 +101,7 @@ Use this table only after the decision table in [`bailian-protocol`](../bailian-
 | Training dataset / fine-tune / deploy                 | → skill `bailian-finetune`                                                                                     | Fallback: `bl dataset\|finetune\|deploy --help`                                                                                                            |
 | agents.yaml IaC / managed-agent sessions              | → skill `bailian-managed-agent`                                                                                | Fallback: `bl managed-agent --help`; `apply`/`destroy` need `--yes` after `plan`                                                                           |
 | Bailian Sandbox instance / template lifecycle         | → skill `bailian-sandbox`                                                                                      | Fallback: `bl sandbox --help`                                                                                                                              |
-| Web search (model-aware routing)                      | → skill `bailian-web-search`                                                                                   | Token Plan vs MCP path + fallback; fallback: `bl search web --help`                                                                                        |
+| Web search (mode-aware routing)                       | → skill `bailian-web-search`                                                                                   | Default model mode vs explicit `--mode mcp` + fallback; fallback: `bl search web --help`                                                                   |
 
 Flags, usage, and examples: see hub [`reference/`](reference/index.md) or `bl <command> --help` — do not guess flags. Domain command details live in the owning skill's `reference/`. Examples in `reference/` stay bare; wrap with Call attribution when running.
 
@@ -149,7 +149,7 @@ schema-export commands.
 ## Routing reminders
 
 - Image/video/audio generation or editing → skill `bailian-gen` (class 3 consent from `bailian-protocol`). Fine-tuning / datasets / deployments → `bailian-finetune`. agents.yaml IaC → `bailian-managed-agent`. Soft hand-off: Read sibling skill if installed; else `bl … --help` or prompt `bl skill init`. Image understanding the host agent can do → host-first; use `bl vision` / `bl omni` only when the user names a Bailian model or the media (video/audio files) exceeds host capability.
-- Web search inside a Bailian workflow → skill `bailian-web-search` (model-aware routing: Token Plan → model-native search; default → `bl search web`; MCP failure → fall back once). Generic web research the host can do stays host-first — do not bounce it through `bl`.
+- Web search inside a Bailian workflow → skill `bailian-web-search` (routing: default `bl search web` model mode for every identity; explicit `--mode mcp` for regular keys; MCP failure → fall back once). Generic web research the host can do stays host-first — do not bounce it through `bl`.
 - Answer ordinary reasoning, coding, writing, translation, summarization, and generic research with the host agent's native capabilities; do not bounce them through `bl text chat` or `bl search web`.
 - Usage / quota / credits questions that do not name a product → ask which product (Bailian or another AI service) first; run `bl usage` / `bl quota` only after the user picks Bailian or Bailian context is already established.
 - Personal recall, new long-term personal facts, and "remember this" requests → read `bailian-memory` if installed to check enablement, including first-use onboarding. Once enabled, use it for authorized personal recall/extraction; installation alone does not authorize cloud uploads. If unavailable or declined, respect the user’s host-memory choice. Explicit Bailian memory resource operations also route to `bailian-memory`.
diff --git a/skills/bailian-cli/reference/index.md b/skills/bailian-cli/reference/index.md
index 1a215a16..2b4b9492 100644
--- a/skills/bailian-cli/reference/index.md
+++ b/skills/bailian-cli/reference/index.md
@@ -100,7 +100,7 @@ Use this index for the skill-scoped quick index and global flags.
 | `bl quota history`              | Console        | View quota change history                                                                                          | [quota.md](quota.md)           |
 | `bl quota list`                 | API Key        | View model rate limits (QPM/TPM, account and workspace level)                                                      | [quota.md](quota.md)           |
 | `bl quota update`               | API Key        | Update model rate limits (QPM/TPM)                                                                                 | [quota.md](quota.md)           |
-| `bl search web`                 | API Key        | Search the web using DashScope MCP WebSearch service                                                               | [search.md](search.md)         |
+| `bl search web`                 | API Key        | Search the web via a chat model with built-in web search (default) or the DashScope MCP WebSearch service          | [search.md](search.md)         |
 | `bl skill add`                  | No Auth        | Install skills from the Bailian skill registry into local agents                                                   | [skill.md](skill.md)           |
 | `bl skill init`                 | No Auth        | Install all bailian-* skills (one-shot bootstrap for new environments)                                             | [skill.md](skill.md)           |
 | `bl skill list`                 | No Auth        | List registry skills and diff against local installs                                                               | [skill.md](skill.md)           |
diff --git a/skills/bailian-cli/reference/search.md b/skills/bailian-cli/reference/search.md
index b311f800..7b63ecb8 100644
--- a/skills/bailian-cli/reference/search.md
+++ b/skills/bailian-cli/reference/search.md
@@ -3,36 +3,42 @@
 > Auto-generated from `packages/cli/src/commands.ts`. Do not edit by hand.
 > Regenerate: `pnpm --filter bailian-cli run generate:reference`.
 
-> **Agent routing (mandatory):** Before running any `bl search *` command, Read skill `bailian-web-search` if installed and follow its identity-based routing (Token Plan → model-native web search; default → MCP; eligible MCP failures → fall back once). Do **not** call `bl search web` directly from this reference alone — Token Plan keys cannot authorize Bailian MCP search. If that skill is missing, run `bl skill init` or fall back to `bl search web --help` / `bl text chat --help` after checking `bl config show --output json`.
+> **Agent routing (mandatory):** Before running any `bl search *` command, Read skill `bailian-web-search` if installed and follow its routing (default `bl search web` = model mode for every identity, including Token Plan; the WebSearch MCP service requires explicit `--mode mcp` and a regular DashScope key; eligible MCP failures → fall back to model mode once). Do **not** route Token Plan keys to `bl search web --mode mcp` — they cannot authorize Bailian MCP search. If that skill is missing, run `bl skill init` or fall back to `bl search web --help` / `bl text chat --help` after checking `bl config show --output json`.
 
 Index: [index.md](index.md)
 
 ## Commands in this group
 
-| Command         | Authentication | Description                                          |
-| --------------- | -------------- | ---------------------------------------------------- |
-| `bl search web` | API Key        | Search the web using DashScope MCP WebSearch service |
+| Command         | Authentication | Description                                                                                               |
+| --------------- | -------------- | --------------------------------------------------------------------------------------------------------- |
+| `bl search web` | API Key        | Search the web via a chat model with built-in web search (default) or the DashScope MCP WebSearch service |
 
 ## Command details
 
 ### `bl search web`
 
-| Field              | Value                                                |
-| ------------------ | ---------------------------------------------------- |
-| **Name**           | `search web`                                         |
-| **Description**    | Search the web using DashScope MCP WebSearch service |
-| **Authentication** | API Key                                              |
-| **Usage**          | `bl search web --query <text> [flags]`               |
+| Field              | Value                                                                                                     |
+| ------------------ | --------------------------------------------------------------------------------------------------------- |
+| **Name**           | `search web`                                                                                              |
+| **Description**    | Search the web via a chat model with built-in web search (default) or the DashScope MCP WebSearch service |
+| **Authentication** | API Key                                                                                                   |
+| **Usage**          | `bl search web --query <text> [flags]`                                                                    |
 
 #### Flags
 
-| Flag               | Type   | Required | Description                            |
-| ------------------ | ------ | -------- | -------------------------------------- |
-| `--query <text>`   | string | no       | Search query text                      |
-| `--count <n>`      | number | no       | Number of search results (default: 10) |
-| `--list-tools`     | switch | no       | List available MCP tools and exit      |
-| `--api-key <key>`  | string | no       | API key                                |
-| `--base-url <url>` | string | no       | API base URL                           |
+| Flag                  | Type   | Required | Description                                                                                                              |
+| --------------------- | ------ | -------- | ------------------------------------------------------------------------------------------------------------------------ |
+| `--query <text>`      | string | no       | Search query text                                                                                                        |
+| `--count <n>`         | number | no       | Number of search results (default: 10)                                                                                   |
+| `--list-tools`        | switch | no       | List available MCP tools and exit                                                                                        |
+| `--mode <mcp\|model>` | string | no       | Search mode: model answers via Chat Completions with enable_search, mcp calls the WebSearch MCP service (default: model) |
+| `--model <model>`     | string | no       | Model ID for --mode model (default: qwen3.8-max)                                                                         |
+| `--api-key <key>`     | string | no       | API key                                                                                                                  |
+| `--base-url <url>`    | string | no       | API base URL                                                                                                             |
+
+#### Notes
+
+- Defaults to --mode model: one non-streaming Chat Completions request with enable_search: true, returning the model answer as a single content[].text block (no pages list). Use --mode mcp for the WebSearch MCP service (pages list); --count and --list-tools are MCP-mode only. The model resolves from --model, default_text_model, then qwen3.8-max.
 
 #### Examples
 
@@ -41,11 +47,15 @@ bl search web --query "Alibaba Cloud Bailian latest features"
 ```
 
 ```bash
-bl search web --query "TypeScript 5.9 new features" --count 5
+bl search web --query "Today's news"
 ```
 
 ```bash
-bl search web --query "Today's news"
+bl search web --mode mcp --query "TypeScript 5.9 new features" --count 5
+```
+
+```bash
+bl search web --mode mcp --query "Alibaba Cloud Bailian latest features"
 ```
 
 ```bash
diff --git a/skills/bailian-cli/reference/text.md b/skills/bailian-cli/reference/text.md
index eadf3d47..87231cf2 100644
--- a/skills/bailian-cli/reference/text.md
+++ b/skills/bailian-cli/reference/text.md
@@ -24,25 +24,26 @@ Index: [index.md](index.md)
 
 #### Flags
 
-| Flag                      | Type   | Required | Description                                                                 |
-| ------------------------- | ------ | -------- | --------------------------------------------------------------------------- |
-| `--api <chat\|responses>` | string | no       | API to call (default: chat)                                                 |
-| `--model <model>`         | string | no       | Model ID (default: qwen3.8-max)                                             |
-| `--message <text>`        | array  | no       | Message text (repeatable, prefix role: to set role); or use --messages-file |
-| `--messages-file <path>`  | string | no       | Messages array for one request (use - for stdin)                            |
-| `--system <text>`         | string | no       | System prompt                                                               |
-| `--max-tokens <n>`        | number | no       | Maximum tokens to generate (default: 4096)                                  |
-| `--temperature <n>`       | number | no       | Sampling temperature (0.0, 2.0]                                             |
-| `--top-p <n>`             | number | no       | Nucleus sampling threshold                                                  |
-| `--stream`                | switch | no       | Stream response tokens (default: on in TTY)                                 |
-| `--tool <json-or-path>`   | array  | no       | Tool definition as JSON or file path (repeatable)                           |
-| `--enable-thinking`       | switch | no       | Enable thinking/reasoning mode (for qwen3/qwq models)                       |
-| `--thinking-budget <n>`   | number | no       | Max tokens for thinking (default: 4096)                                     |
-| `--batch`                 | switch | no       | Single synchronous Batch Chat request (non-streaming)                       |
-| `--prime`                 | switch | no       | Use Prime mode with a workspace-scoped endpoint                             |
-| `--workspace-id <id>`     | string | no       | Workspace ID for the default Prime endpoint (or set BAILIAN_WORKSPACE_ID)   |
-| `--api-key <key>`         | string | no       | API key                                                                     |
-| `--base-url <url>`        | string | no       | API base URL                                                                |
+| Flag                            | Type    | Required | Description                                                                           |
+| ------------------------------- | ------- | -------- | ------------------------------------------------------------------------------------- |
+| `--api <chat\|responses>`       | string  | no       | API to call (default: chat)                                                           |
+| `--model <model>`               | string  | no       | Model ID (default: qwen3.8-max)                                                       |
+| `--message <text>`              | array   | no       | Message text (repeatable, prefix role: to set role); or use --messages-file           |
+| `--messages-file <path>`        | string  | no       | Messages array for one request (use - for stdin)                                      |
+| `--system <text>`               | string  | no       | System prompt                                                                         |
+| `--max-tokens <n>`              | number  | no       | Maximum tokens to generate (default: 4096)                                            |
+| `--temperature <n>`             | number  | no       | Sampling temperature (0.0, 2.0]                                                       |
+| `--top-p <n>`                   | number  | no       | Nucleus sampling threshold                                                            |
+| `--enable-search <true\|false>` | boolean | no       | Enable built-in web search (Chat Completions API only; default: true for qwen3.8-max) |
+| `--stream`                      | switch  | no       | Stream response tokens (default: on in TTY)                                           |
+| `--tool <json-or-path>`         | array   | no       | Tool definition as JSON or file path (repeatable)                                     |
+| `--enable-thinking`             | switch  | no       | Enable thinking/reasoning mode (for qwen3/qwq models)                                 |
+| `--thinking-budget <n>`         | number  | no       | Max tokens for thinking (default: 4096)                                               |
+| `--batch`                       | switch  | no       | Single synchronous Batch Chat request (non-streaming)                                 |
+| `--prime`                       | switch  | no       | Use Prime mode with a workspace-scoped endpoint                                       |
+| `--workspace-id <id>`           | string  | no       | Workspace ID for the default Prime endpoint (or set BAILIAN_WORKSPACE_ID)             |
+| `--api-key <key>`               | string  | no       | API key                                                                               |
+| `--base-url <url>`              | string  | no       | API base URL                                                                          |
 
 #### Notes
 
@@ -50,6 +51,7 @@ Index: [index.md](index.md)
 - Batch Chat total request timeout defaults to 300s; --timeout must be 60–3600s.
 - Prime mode requires an explicit model and currently supports only the Chat Completions API.
 - Prime endpoint: configured --base-url, DASHSCOPE_BASE_URL, or profile base_url takes precedence; otherwise workspace is resolved from --workspace-id, BAILIAN_WORKSPACE_ID, then config workspace_id.
+- --enable-search maps to the Chat Completions enable_search parameter and requires a model with built-in web search support; whitelisted models (qwen3.8-max) enable it by default unless --enable-search false is passed; for the Responses API use --tool '{"type":"web_search"}' instead.
 
 #### Examples
 
@@ -89,6 +91,10 @@ bl text chat --message "Hello" --output json
 bl text chat --model qwq-plus --message "Solve 1+1" --enable-thinking
 ```
 
+```bash
+bl text chat --enable-search true --message "Latest news about Alibaba Cloud"
+```
+
 ```bash
 bl text chat --prime --workspace-id <id> --model glm-5.3-prime --message "Explain this code"
 ```
diff --git a/skills/bailian-finetune/SKILL.md b/skills/bailian-finetune/SKILL.md
index 81fc32fb..b881ffbf 100644
--- a/skills/bailian-finetune/SKILL.md
+++ b/skills/bailian-finetune/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: bailian-finetune
 metadata:
-  version: "2.8.2"
+  version: "2.8.3"
   requires:
     bins: ["bl"]
 description: >-
diff --git a/skills/bailian-gen/SKILL.md b/skills/bailian-gen/SKILL.md
index e79d9bc9..1189a019 100644
--- a/skills/bailian-gen/SKILL.md
+++ b/skills/bailian-gen/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: bailian-gen
 metadata:
-  version: "2.8.2"
+  version: "2.8.3"
   requires:
     bins: ["bl"]
 description: >-
diff --git a/skills/bailian-knowledge/SKILL.md b/skills/bailian-knowledge/SKILL.md
index 677c8da3..b47c6e05 100644
--- a/skills/bailian-knowledge/SKILL.md
+++ b/skills/bailian-knowledge/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: bailian-knowledge
 metadata:
-  version: "2.8.2"
+  version: "2.8.3"
   requires:
     bins: ["bl"]
 description: >-
diff --git a/skills/bailian-managed-agent/SKILL.md b/skills/bailian-managed-agent/SKILL.md
index df291f26..e9de66db 100644
--- a/skills/bailian-managed-agent/SKILL.md
+++ b/skills/bailian-managed-agent/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: bailian-managed-agent
 metadata:
-  version: "2.8.2"
+  version: "2.8.3"
   requires:
     bins: ["bl"]
 description: >-
diff --git a/skills/bailian-memory/SKILL.md b/skills/bailian-memory/SKILL.md
index f9a5b9cc..d95ec262 100644
--- a/skills/bailian-memory/SKILL.md
+++ b/skills/bailian-memory/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: bailian-memory
 metadata:
-  version: "2.8.2"
+  version: "2.8.3"
   requires:
     bins: ["bl"]
 description: >-
diff --git a/skills/bailian-protocol/SKILL.md b/skills/bailian-protocol/SKILL.md
index 896754ef..a113094e 100644
--- a/skills/bailian-protocol/SKILL.md
+++ b/skills/bailian-protocol/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: bailian-protocol
 metadata:
-  version: "2.8.2"
+  version: "2.8.3"
   requires:
     bins: ["bl"]
 description: >-
diff --git a/skills/bailian-sandbox/SKILL.md b/skills/bailian-sandbox/SKILL.md
index ac5cd047..43e0df22 100644
--- a/skills/bailian-sandbox/SKILL.md
+++ b/skills/bailian-sandbox/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: bailian-sandbox
 metadata:
-  version: "2.8.2"
+  version: "2.8.3"
   requires:
     bins: ["bl"]
 description: >-
diff --git a/skills/bailian-web-search/SKILL.md b/skills/bailian-web-search/SKILL.md
index 4793367a..9f49f1a6 100644
--- a/skills/bailian-web-search/SKILL.md
+++ b/skills/bailian-web-search/SKILL.md
@@ -1,16 +1,15 @@
 ---
 name: bailian-web-search
 metadata:
-  version: "2.8.2"
+  version: "2.8.3"
   requires:
     bins: ["bl"]
 description: >-
   阿里云百炼联网搜索（web search）入口：为联网搜索 / 网页搜索 / 查最新资讯做路径分发。
-  先识别当前连接身份：Token Plan（profile `token-plan` 或 base_url host 为
-  `token-plan.<region>.maas.aliyuncs.com`）→ 模型自带搜索
-  （`bl text chat --api responses --tool '{"type":"web_search"}'`）；其他 / 默认 →
-  Bailian MCP（`bl search web`）；仅在 MCP 鉴权失败、未开通或传输失败时兜底一次到模型自带搜索。
-  Token Plan 联网搜索失败、两条路径鉴权混淆的排查也走本技能。
+  `bl search web` 默认走 model 模式（Chat Completions enable_search），任何连接身份
+  （含 Token Plan）均可用；仅当用户明确要 WebSearch MCP 服务或 pages 列表时用
+  `--mode mcp`（仅普通 DashScope API Key）；MCP 鉴权失败、未开通或传输失败时
+  兜底一次到 model 模式。联网搜索失败、两条路径鉴权混淆的排查也走本技能。
   反触发：宿主可完成的 普通问答 / 编程 / 写作 → 不触发；知识库 RAG → bailian-cli
   （`bl knowledge`）；生图/生视频/语音 → bailian-gen；精调/训练 → bailian-finetune；
   agents.yaml → bailian-managed-agent。
@@ -32,7 +31,7 @@ BAILIAN_CALLER_TYPE=agent BAILIAN_CALLER_NAME=codex BAILIAN_SKILL_NAME=bailian-w
 >
 > **Install (supported):** `bl skill init`
 
-Token Plan keys do not authorize Bailian MCP search — never route them to `bl search web`.
+Token Plan keys do not authorize Bailian MCP search — never route them to `bl search web --mode mcp` (the default model mode works for Token Plan).
 
 ## Routing workflow
 
@@ -51,50 +50,63 @@ Treat the connection as **Token Plan** when **either** condition holds — this
 
 Do not rely on the profile name alone: users can create additional Token Plan profiles under custom names, and only the Base URL host check catches those. Anything else → **default** identity.
 
-### Step 2 — Route by identity
+### Step 2 — Route by mode and identity
 
-| Condition                 | Route                           | Notes                                               |
-| ------------------------- | ------------------------------- | --------------------------------------------------- |
-| Token Plan model identity | Model-native path               | Preferred — the only path Token Plan keys authorize |
-| Any other model / default | MCP path                        | Default for regular DashScope API keys              |
-| MCP path (eligible fail)  | Model-native path, exactly once | Fallback — only for the failure classes below       |
+| Condition                                     | Route                                 | Notes                                               |
+| --------------------------------------------- | ------------------------------------- | --------------------------------------------------- |
+| Any identity (default)                        | Model mode: `bl search web --query …` | CLI default; the model answers with `enable_search` |
+| Regular key + user explicitly wants MCP/pages | `bl search web --mode mcp --query …`  | Requires the WebSearch MCP activated for that key   |
+| Token Plan + user asks for MCP                | Explain, stay on model mode           | Token Plan keys cannot authorize Bailian MCP search |
+| `--mode mcp` run (eligible fail)              | Model mode, exactly once              | Fallback — only for the failure classes below       |
 
-**User override:** if the user names a specific path or command, follow it **except** under Token Plan identity: do not call `bl search web` even if the user asks for MCP — explain that Token Plan keys cannot authorize Bailian MCP search, then use the model-native path (or ask once whether to switch profile / key).
+**User override:** if the user names a specific path or command, follow it **except** under Token Plan identity: do not call `bl search web --mode mcp` even if the user asks for MCP — explain that Token Plan keys cannot authorize Bailian MCP search, then use the default model mode (or ask once whether to switch profile / key).
 
-### Model-native path (Token Plan preferred / fallback target)
+### Model path (default for every identity)
+
+```bash
+bl search web --query "阿里云百炼最新功能"
+```
+
+- Without `--mode` the CLI runs model mode: one non-streaming Chat Completions request with `enable_search: true`; the reply is the model answer (no pages list).
+- Works for Token Plan and regular identities alike: the CLI selects the credential per mode (model mode authenticates in the `text.chat` capability domain).
+- **Do not hardcode `--model`:** omit it so the CLI uses `default_text_model` / built-in default; only pass `--model` when the user named one.
+- Flags / usage: Read skill `bailian-cli` if installed, else `bl search web --help`.
+
+### Responses API alternative (only when the user names it)
 
 ```bash
 bl text chat --api responses --tool '{"type":"web_search"}' --message "搜索近期的阿里云新闻"
 ```
 
-- The Responses API enables native web search via the tool definition `{"type":"web_search"}`.
-- Requires a model with native web search support (Qwen3.7+). **Do not hardcode `--model`:** omit it so the CLI uses `default_text_model` / built-in default; only pass `--model` when the user named one.
+- The Responses API enables native web search via the tool definition `{"type":"web_search"}`; prefer it only when the user explicitly asks for the Responses API or tool-style results.
+- Requires a model with native web search support (Qwen3.7+); the same "do not hardcode `--model`" rule applies.
 - Write `--message` in the user's language; the reply language follows the prompt (see `bailian-protocol` → Respond in the user's language).
 - Flags / usage: Read skill `bailian-cli` if installed, else `bl text chat --help`.
 
-### MCP path (default)
+### MCP path (explicit `--mode mcp`, regular keys only)
 
 ```bash
-bl search web --query "阿里云百炼最新功能"
+bl search web --mode mcp --query "阿里云百炼最新功能"
 ```
 
+- Route here only when the user explicitly wants the WebSearch MCP service or the pages list; regular DashScope API keys only — Token Plan keys cannot authorize it.
 - Requires the WebSearch MCP to be activated for the current key; on the not-activated error the CLI appends an activation hint with the marketplace URL — relay it to the user.
 - Flags / usage: Read skill `bailian-cli` if installed, else `bl search web --help`.
 
-### Fallback (MCP → model-native, exactly once)
+### Fallback (MCP → model, exactly once)
 
-**Do not** fall back on every non-zero exit. Fall back **only** when `bl search web` fails for one of these classes (match stderr / message):
+**Do not** fall back on every non-zero exit. Fall back **only** when a `bl search web --mode mcp` run fails for one of these classes (match stderr / message):
 
-- **auth / permission** — key not valid for the MCP service (e.g. Token Plan key misrouted by identity detection),
+- **auth / permission** — key not valid for the MCP service (e.g. Token Plan key misrouted to `--mode mcp`),
 - **MCP not activated** — `MCP request failed: 404` with `未开通` / `MCP不存在` / `MCP_IS_INVALID` (CLI may append an activation hint),
 - **MCP transport** — 405 / Streamable-HTTP unsupported, or clear network / timeout / DNS failures reaching the MCP endpoint.
 
-**Do not fall back** for: missing `--query` / USAGE errors, rate limits, content-policy / business errors from a successful MCP session, or empty-but-successful result sets. Report those verbatim and stop (or ask the user); do not burn a Responses call.
+**Do not fall back** for: missing `--query` / USAGE errors, rate limits, content-policy / business errors from a successful MCP session, or empty-but-successful result sets. Report those verbatim and stop (or ask the user); do not burn a model call. Default model-mode failures are **not** MCP failures — report them verbatim; never reinterpret them as MCP errors or retry them as MCP.
 
 Fallback discipline:
 
-1. Re-issue the same query via the model-native path (omit `--model` unless the user named one).
-2. If the fallback succeeds, tell the user the MCP path failed and — when the not-activated hint appeared — that activating the WebSearch MCP restores the default path.
+1. Re-issue the same query via the default model mode (omit `--model` unless the user named one).
+2. If the fallback succeeds, tell the user the MCP path failed and — when the not-activated hint appeared — that activating the WebSearch MCP restores the MCP path.
 3. If the fallback also fails, stop and report both errors verbatim; follow the issue-reporting flow in `bailian-protocol` (ask once). Never loop retries.
 
 ## Quick examples
@@ -103,21 +115,21 @@ Fallback discipline:
 # Step 1: identify the active model identity
 bl config show --output json
 
-# Token Plan identity → model-native web search (no --model unless user named one)
-bl text chat --api responses --tool '{"type":"web_search"}' --message "搜索近期的阿里云新闻"
-
-# Default identity → Bailian MCP search
+# Default (any identity, including Token Plan): model mode
 bl search web --query "阿里云百炼最新功能"
 
-# Eligible MCP failure → fall back once (same rule: no hardcoded --model)
-bl text chat --api responses --tool '{"type":"web_search"}' --message "阿里云百炼最新功能"
+# Regular key + user explicitly wants the MCP service / pages list
+bl search web --mode mcp --query "阿里云百炼最新功能"
+
+# Responses API alternative (only when the user names it)
+bl text chat --api responses --tool '{"type":"web_search"}' --message "搜索近期的阿里云新闻"
 ```
 
 ## Routing reminders
 
 - Generic web research the host can do, ordinary Q&A, coding, writing → host-first; do not invoke `bl` (class 1 in `bailian-protocol`). Route only when the user names Bailian / DashScope / `bl` or continues an existing `bl` workflow (class 4).
 - Knowledge-base / RAG over Bailian corpora → hub skill `bailian-cli` (`bl knowledge`), not this skill.
-- Summarize search results in the user's language; on the model-native path the CLI injects no default language — if a fixed language is required, pass `--system` in that language (do not hardcode 简体中文).
+- Summarize search results in the user's language; write `--query` / `--message` in the user's language — the model path injects no default language (do not hardcode 简体中文); if a fixed reply language is required, use the Responses API alternative with `--system` in that language.
 - Other Bailian workflows (apps / usage / config) → hub skill `bailian-cli`; media generation → `bailian-gen`; fine-tuning → `bailian-finetune`; agents.yaml → `bailian-managed-agent`. Soft hand-off by skill name: Read if installed, else `bl … --help` or prompt `bl skill init`.
 
 ## references
diff --git a/tools/generate-reference.ts b/tools/generate-reference.ts
index bcb6e373..a194427a 100644
--- a/tools/generate-reference.ts
+++ b/tools/generate-reference.ts
@@ -76,9 +76,11 @@ const GENERATED_BANNER =
 const GROUP_AGENT_NOTICE: Readonly<Record<string, string>> = {
   search:
     "> **Agent routing (mandatory):** Before running any `bl search *` command, " +
-    "Read skill `bailian-web-search` if installed and follow its identity-based routing " +
-    "(Token Plan → model-native web search; default → MCP; eligible MCP failures → fall back once). " +
-    "Do **not** call `bl search web` directly from this reference alone — Token Plan keys cannot " +
+    "Read skill `bailian-web-search` if installed and follow its routing " +
+    "(default `bl search web` = model mode for every identity, including Token Plan; " +
+    "the WebSearch MCP service requires explicit `--mode mcp` and a regular DashScope key; " +
+    "eligible MCP failures → fall back to model mode once). " +
+    "Do **not** route Token Plan keys to `bl search web --mode mcp` — they cannot " +
     "authorize Bailian MCP search. If that skill is missing, run `bl skill init` or fall back to " +
     "`bl search web --help` / `bl text chat --help` after checking `bl config show --output json`.",
 };