feat(zai): add z.ai (GLM-4.6) as a new LLM provider

packages/browseros-agent/apps/agent/entrypoints/app/agents/openclaw-supported-providers.ts

@@ -6,6 +6,7 @@ const OPENCLAW_SUPPORTED_PROVIDER_TYPES: ProviderType[] = [
   'openai-compatible',
   'anthropic',
   'moonshot',
+  'zai',
 ]
 
 export function isOpenClawSupportedProviderType(

packages/browseros-agent/apps/agent/entrypoints/app/ai-settings/NewProviderDialog.tsx

@@ -87,6 +87,7 @@ const providerTypeEnum = z.enum([
   'chatgpt-pro',
   'github-copilot',
   'qwen-code',
+  'zai',
 ])
 
 /**

packages/browseros-agent/apps/agent/lib/browseros/helpers.ts

@@ -21,6 +21,11 @@ export class McpPortError extends Error {
  * @public
  */
 export async function getAgentServerUrl(): Promise<string> {
+  // In development, always use the configured server port directly
+  if (env.VITE_BROWSEROS_SERVER_PORT) {
+    return `http://127.0.0.1:${env.VITE_BROWSEROS_SERVER_PORT}`
+  }
+
   const supportsUnifiedPort = await Capabilities.supports(
     Feature.UNIFIED_PORT_SUPPORT,
   )

packages/browseros-agent/apps/agent/lib/llm-providers/models-dev-data.json

@@ -5402,5 +5402,45 @@
         "outputCost": 0
       }
     ]
+  },
+  "zai": {
+    "name": "z.ai",
+    "api": "https://api.z.ai/api/coding/paas/v4",
+    "doc": "https://docs.z.ai/guides/llm/glm-4.6",
+    "models": [
+      {
+        "id": "glm-4.6",
+        "name": "GLM-4.6",
+        "contextWindow": 200000,
+        "maxOutput": 131072,
+        "supportsImages": true,
+        "supportsReasoning": true,
+        "supportsToolCall": true,
+        "inputCost": 0.6,
+        "outputCost": 2.2
+      },
+      {
+        "id": "glm-4.5",
+        "name": "GLM-4.5",
+        "contextWindow": 128000,
+        "maxOutput": 96000,
+        "supportsImages": true,
+        "supportsReasoning": true,
+        "supportsToolCall": true,
+        "inputCost": 0.6,
+        "outputCost": 2.2
+      },
+      {
+        "id": "glm-4.5-air",
+        "name": "GLM-4.5-Air",
+        "contextWindow": 128000,
+        "maxOutput": 96000,
+        "supportsImages": true,
+        "supportsReasoning": true,
+        "supportsToolCall": true,
+        "inputCost": 0.2,
+        "outputCost": 1.1
+      }
+    ]
   }
 }

packages/browseros-agent/apps/agent/lib/llm-providers/providerIcons.tsx

@@ -9,6 +9,7 @@ import {
   OpenAI,
   OpenRouter,
   Qwen,
+  ZAI,
 } from '@lobehub/icons'
 import { Bot, Github } from 'lucide-react'
 import type { FC, SVGProps } from 'react'
@@ -36,6 +37,7 @@ const providerIconMap: Record<ProviderType, IconComponent | null> = {
   'chatgpt-pro': OpenAI,
   'github-copilot': Github,
   'qwen-code': Qwen,
+  zai: ZAI,
 }
 
 interface ProviderIconProps {

packages/browseros-agent/apps/agent/lib/llm-providers/providerTemplates.ts

@@ -140,6 +140,16 @@ export const providerTemplates: ProviderTemplate[] = [
     setupGuideUrl:
       'https://docs.aws.amazon.com/bedrock/latest/userguide/getting-started.html',
   }),
+  {
+    id: 'zai',
+    name: 'z.ai',
+    defaultBaseUrl: 'https://api.z.ai/api/coding/paas/v4',
+    defaultModelId: 'glm-4.6',
+    supportsImages: true,
+    contextWindow: 200000,
+    apiKeyUrl: 'https://z.ai/manage-apikey/apikey-list',
+    setupGuideUrl: 'https://docs.z.ai/guides/llm/glm-4.6',
+  },
 ]
 
 /**
@@ -161,6 +171,7 @@ export const providerTypeOptions: { value: ProviderType; label: string }[] = [
   { value: 'lmstudio', label: 'LM Studio' },
   { value: 'bedrock', label: 'AWS Bedrock' },
   { value: 'browseros', label: 'BrowserOS' },
+  { value: 'zai', label: 'z.ai' },
 ]
 
 /**
@@ -192,6 +203,7 @@ export const DEFAULT_BASE_URLS: Record<ProviderType, string> = {
   lmstudio: 'http://localhost:1234/v1',
   bedrock: '',
   browseros: '',
+  zai: 'https://api.z.ai/api/coding/paas/v4',
 }
 
 /**

packages/browseros-agent/apps/agent/lib/llm-providers/types.ts

@@ -17,6 +17,7 @@ export type ProviderType =
   | 'chatgpt-pro'
   | 'github-copilot'
   | 'qwen-code'
+  | 'zai'
 
 /**
  * LLM Provider configuration

packages/browseros-agent/apps/agent/web-ext.config.ts

@@ -7,7 +7,6 @@ const chromiumArgs = [
   '--use-mock-keychain',
   '--show-component-extension-options',
   '--disable-browseros-server',
-  '--disable-browseros-extensions',
   '--browseros-dock-icon=dev',
 ]
 

packages/browseros-agent/apps/server/src/agent/provider-factory.ts

@@ -168,6 +168,17 @@ function createMoonshotFactory(
   })
 }
 
+function createZaiFactory(
+  config: ResolvedAgentConfig,
+): (modelId: string) => unknown {
+  if (!config.apiKey) throw new Error('z.ai provider requires apiKey')
+  return createOpenAICompatible({
+    name: 'zai',
+    baseURL: config.baseUrl || EXTERNAL_URLS.ZAI_API,
+    apiKey: config.apiKey,
+  })
+}
+
 function createQwenCodeFactory(
   config: ResolvedAgentConfig,
 ): (modelId: string) => unknown {
@@ -218,6 +229,7 @@ const PROVIDER_FACTORIES: Record<string, ProviderFactory> = {
   [LLM_PROVIDERS.CHATGPT_PRO]: createChatGPTProFactory,
   [LLM_PROVIDERS.GITHUB_COPILOT]: createGitHubCopilotFactory,
   [LLM_PROVIDERS.QWEN_CODE]: createQwenCodeFactory,
+  [LLM_PROVIDERS.ZAI]: createZaiFactory,
 }
 
 export function createLanguageModel(

packages/browseros-agent/apps/server/src/api/routes/provider.ts

@@ -26,6 +26,13 @@ export function createProviderRoutes(deps: ProviderRouteDeps = {}) {
         model: config.model,
       })
 
+      logger.info('Testing provider connection start', {
+        provider: config.provider,
+        model: config.model,
+        baseUrl: config.baseUrl,
+        hasApiKey: !!config.apiKey,
+      })
+
       const result = await testProviderConnection(config, deps.browserosId)
 
       logger.info('Provider test result', {

packages/browseros-agent/apps/server/src/api/services/openclaw/openclaw-provider-map.ts

@@ -9,6 +9,7 @@ export const SUPPORTED_OPENCLAW_PROVIDERS = [
   'openai',
   'anthropic',
   'moonshot',
+  'zai',
 ] as const
 
 export type SupportedOpenClawProvider =
@@ -32,6 +33,7 @@ const PROVIDER_ENV_VARS: Record<SupportedOpenClawProvider, string> = {
   moonshot: 'MOONSHOT_API_KEY',
   openai: 'OPENAI_API_KEY',
   openrouter: 'OPENROUTER_API_KEY',
+  zai: 'ZAI_API_KEY',
 }
 
 export class UnsupportedOpenClawProviderError extends Error {

packages/browseros-agent/apps/server/src/browser/browser.ts

@@ -188,10 +188,43 @@ export class Browser {
   // --- Pages ---
 
   async listPages(): Promise<PageInfo[]> {
-    const result = await this.cdp.Browser.getTabs({ includeHidden: true })
-    const tabs = (result.tabs as TabInfo[]).filter(
-      (t) => !EXCLUDED_URL_PREFIXES.some((prefix) => t.url.startsWith(prefix)),
-    )
+    let tabs: TabInfo[]
+    try {
+      const result = await this.cdp.Browser.getTabs({ includeHidden: true })
+      tabs = (result.tabs as TabInfo[]).filter(
+        (t) =>
+          !EXCLUDED_URL_PREFIXES.some((prefix) => t.url.startsWith(prefix)),
+      )
+    } catch (err) {
+      const msg = err instanceof Error ? err.message : String(err)
+      // Fallback for CDP implementations without Browser.getTabs (e.g. some
+      // Chromium forks or older builds).
+      if (msg.includes("wasn't found") || msg.includes('not found')) {
+        const result = await this.cdp.Target.getTargets()
+        tabs = result.targetInfos
+          .filter((t) => t.type === 'page')
+          .map((t) => ({
+            tabId: t.tabId ?? 0,
+            targetId: t.targetId,
+            url: t.url,
+            title: t.title,
+            isActive: false,
+            isLoading: false,
+            loadProgress: 1,
+            isPinned: false,
+            isHidden: false,
+            windowId: t.windowId,
+            index: 0,
+            groupId: undefined,
+          }))
+          .filter(
+            (t) =>
+              !EXCLUDED_URL_PREFIXES.some((prefix) => t.url.startsWith(prefix)),
+          )
+      } else {
+        throw err
+      }
+    }
 
     const seenTargetIds = new Set<string>()
 

packages/browseros-agent/apps/server/src/lib/clients/llm/provider.ts

@@ -152,6 +152,20 @@ function createMoonshotModel(config: ResolvedLLMConfig): LanguageModel {
   })(config.model)
 }
 
+function createZaiModel(config: ResolvedLLMConfig): LanguageModel {
+  logger.info('createZaiModel', {
+    model: config.model,
+    baseUrl: config.baseUrl || EXTERNAL_URLS.ZAI_API,
+    hasApiKey: !!config.apiKey,
+  })
+  if (!config.apiKey) throw new Error('z.ai provider requires apiKey')
+  return createOpenAICompatible({
+    name: 'zai',
+    baseURL: config.baseUrl || EXTERNAL_URLS.ZAI_API,
+    apiKey: config.apiKey,
+  })(config.model)
+}
+
 function createQwenCodeModel(config: ResolvedLLMConfig): LanguageModel {
   if (!config.apiKey) throw new Error('Qwen Code requires OAuth authentication')
   return createOpenAICompatible({
@@ -196,6 +210,7 @@ const PROVIDER_FACTORIES: Record<string, ProviderFactory> = {
   [LLM_PROVIDERS.CHATGPT_PRO]: createChatGPTProModel,
   [LLM_PROVIDERS.GITHUB_COPILOT]: createGitHubCopilotModel,
   [LLM_PROVIDERS.QWEN_CODE]: createQwenCodeModel,
+  [LLM_PROVIDERS.ZAI]: createZaiModel,
 }
 
 export function createLLMProvider(config: ResolvedLLMConfig): LanguageModel {

packages/browseros-agent/apps/server/src/lib/clients/llm/test-provider.ts

@@ -6,7 +6,8 @@
 
 import { TIMEOUTS } from '@browseros/shared/constants/timeouts'
 import type { LLMConfig } from '@browseros/shared/schemas/llm'
-import { streamText } from 'ai'
+import { generateText } from 'ai'
+import { logger } from '../../logger'
 import { resolveLLMConfig } from './config'
 import { createLLMProvider } from './provider'
 
@@ -30,18 +31,34 @@ export async function testProviderConnection(
   const startTime = performance.now()
 
   try {
+    logger.info('testProviderConnection start', {
+      provider: config.provider,
+      model: config.model,
+      baseUrl: config.baseUrl,
+      hasApiKey: !!config.apiKey,
+    })
     const resolvedConfig = await resolveLLMConfig(config, browserosId)
+    logger.info('testProviderConnection resolved', {
+      provider: resolvedConfig.provider,
+      model: resolvedConfig.model,
+      baseUrl: resolvedConfig.baseUrl,
+    })
     const model = createLLMProvider(resolvedConfig)
+    logger.info('testProviderConnection model created', {
+      provider: resolvedConfig.provider,
+    })
 
-    // streamText works for all providers including Codex (which requires streaming)
-    const stream = streamText({
+    // Use generateText for testing to get clear API errors (streamText wraps
+    // APICallError in NoOutputGeneratedError and loses responseBody details).
+    const result = await generateText({
       model,
       messages: [{ role: 'user', content: TEST_PROMPT }],
+      maxRetries: 0,
       abortSignal: AbortSignal.timeout(TIMEOUTS.TEST_PROVIDER),
     })
-    const text = await stream.text
     const responseTime = Math.round(performance.now() - startTime)
 
+    const text = result.text
     if (text) {
       const preview = text.length > 100 ? `${text.slice(0, 100)}...` : text
       return {
@@ -58,7 +75,13 @@ export async function testProviderConnection(
     }
   } catch (error) {
     const responseTime = Math.round(performance.now() - startTime)
-    const errorMessage = error instanceof Error ? error.message : String(error)
+    logger.info('testProviderConnection caught error', {
+      provider: config.provider,
+      errorType: typeof error,
+      errorName: error instanceof Error ? error.name : undefined,
+      errorMessage: error instanceof Error ? error.message : String(error),
+    })
+    const errorMessage = extractProviderErrorMessage(error, config.provider)
 
     return {
       success: false,
@@ -67,3 +90,37 @@ export async function testProviderConnection(
     }
   }
 }
+
+function extractProviderErrorMessage(
+  error: unknown,
+  _provider: string,
+): string {
+  // Check for API call error with response body (generateText preserves
+  // APICallError directly, so responseBody is available on the error object)
+  if (
+    error != null &&
+    typeof error === 'object' &&
+    'responseBody' in error &&
+    typeof (error as { responseBody?: string }).responseBody === 'string'
+  ) {
+    try {
+      const parsed = JSON.parse(
+        (error as { responseBody: string }).responseBody,
+      )
+      const msg =
+        parsed?.error?.message ||
+        parsed?.message ||
+        parsed?.error?.code ||
+        (error instanceof Error ? error.message : String(error))
+      return msg
+    } catch {
+      // Not valid JSON, fall through
+    }
+  }
+
+  if (error instanceof Error) {
+    return error.message
+  }
+
+  return String(error)
+}

packages/browseros-agent/packages/shared/src/constants/urls.ts

@@ -19,4 +19,5 @@ export const EXTERNAL_URLS = {
   QWEN_DEVICE_CODE: 'https://chat.qwen.ai/api/v1/oauth2/device/code',
   QWEN_OAUTH_TOKEN: 'https://chat.qwen.ai/api/v1/oauth2/token',
   QWEN_CODE_API: 'https://portal.qwen.ai/v1',
+  ZAI_API: 'https://api.z.ai/api/coding/paas/v4',
 } as const