diff --git a/.claude/cli-settings.json b/.claude/cli-settings.json
new file mode 100644
index 00000000..3d4b8010
--- /dev/null
+++ b/.claude/cli-settings.json
@@ -0,0 +1,18 @@
+{
+  "version": "1.0.0",
+  "defaultTool": "gemini",
+  "promptFormat": "plain",
+  "smartContext": {
+    "enabled": false,
+    "maxFiles": 10
+  },
+  "nativeResume": true,
+  "recursiveQuery": true,
+  "cache": {
+    "injectionMode": "auto",
+    "defaultPrefix": "",
+    "defaultSuffix": ""
+  },
+  "codeIndexMcp": "ace",
+  "$schema": "./cli-settings.schema.json"
+}
diff --git a/.claude/cli-tools.json b/.claude/cli-tools.json
index 412156ad..a644a7e3 100644
--- a/.claude/cli-tools.json
+++ b/.claude/cli-tools.json
@@ -1,39 +1,50 @@
 {
-  "version": "1.0.0",
+  "version": "2.0.0",
   "tools": {
     "gemini": {
       "enabled": true,
       "isBuiltin": true,
       "command": "gemini",
-      "description": "Google AI for code analysis"
+      "description": "Google AI for code analysis",
+      "tags": []
     },
     "qwen": {
       "enabled": true,
       "isBuiltin": true,
       "command": "qwen",
-      "description": "Alibaba AI assistant"
+      "description": "Alibaba AI assistant",
+      "tags": []
     },
     "codex": {
       "enabled": true,
       "isBuiltin": true,
       "command": "codex",
-      "description": "OpenAI code generation"
+      "description": "OpenAI code generation",
+      "tags": []
     },
     "claude": {
       "enabled": true,
       "isBuiltin": true,
       "command": "claude",
-      "description": "Anthropic AI assistant"
+      "description": "Anthropic AI assistant",
+      "tags": []
     },
     "opencode": {
       "enabled": true,
       "isBuiltin": true,
       "command": "opencode",
       "description": "OpenCode AI assistant",
-      "primaryModel": "opencode/glm-4.7-free"
+      "primaryModel": "opencode/glm-4.7-free",
+      "tags": []
     }
   },
-  "customEndpoints": [],
+  "customEndpoints": [
+    {
+      "id": "g25",
+      "name": "g25",
+      "enabled": true
+    }
+  ],
   "defaultTool": "gemini",
   "settings": {
     "promptFormat": "plain",
@@ -48,7 +59,7 @@
       "defaultPrefix": "",
       "defaultSuffix": ""
     },
-    "codeIndexMcp": "ace"
+    "codeIndexMcp": "codexlens"
   },
   "$schema": "./cli-tools.schema.json"
 }
\ No newline at end of file
diff --git a/ccw/src/config/litellm-api-config-manager.ts b/ccw/src/config/litellm-api-config-manager.ts
index a5b12f20..a7ec0d54 100644
--- a/ccw/src/config/litellm-api-config-manager.ts
+++ b/ccw/src/config/litellm-api-config-manager.ts
@@ -1033,5 +1033,219 @@ function objectToYaml(obj: unknown, indent: number = 0): string {
   return String(obj);
 }
 
+// ===========================
+// Multi-Model Pool Management
+// ===========================
+
+/**
+ * Migrate legacy embeddingPoolConfig to new modelPools array
+ * This function ensures backward compatibility with existing configurations
+ */
+function migrateEmbeddingPoolToModelPools(config: LiteLLMApiConfig): void {
+  // Skip if already has modelPools or no legacy config
+  if (config.modelPools && config.modelPools.length > 0) return;
+  if (!config.embeddingPoolConfig) return;
+
+  // Convert legacy embeddingPoolConfig to ModelPoolConfig
+  const legacyPool = config.embeddingPoolConfig;
+  const modelPool: import('../types/litellm-api-config.js').ModelPoolConfig = {
+    id: `pool-embedding-${Date.now()}`,
+    modelType: 'embedding',
+    enabled: legacyPool.enabled,
+    targetModel: legacyPool.targetModel,
+    strategy: legacyPool.strategy,
+    autoDiscover: legacyPool.autoDiscover,
+    excludedProviderIds: legacyPool.excludedProviderIds || [],
+    defaultCooldown: legacyPool.defaultCooldown,
+    defaultMaxConcurrentPerKey: legacyPool.defaultMaxConcurrentPerKey,
+    name: `Embedding Pool - ${legacyPool.targetModel}`,
+    description: 'Migrated from legacy embeddingPoolConfig',
+  };
+
+  config.modelPools = [modelPool];
+  // Keep legacy config for backward compatibility with old CodexLens versions
+}
+
+/**
+ * Get all model pool configurations
+ * Returns empty array if no pools configured
+ */
+export function getModelPools(baseDir: string): import('../types/litellm-api-config.js').ModelPoolConfig[] {
+  const config = loadLiteLLMApiConfig(baseDir);
+  
+  // Auto-migrate if needed
+  migrateEmbeddingPoolToModelPools(config);
+  
+  return config.modelPools || [];
+}
+
+/**
+ * Get a specific model pool by ID
+ */
+export function getModelPool(
+  baseDir: string,
+  poolId: string
+): import('../types/litellm-api-config.js').ModelPoolConfig | undefined {
+  const pools = getModelPools(baseDir);
+  return pools.find(p => p.id === poolId);
+}
+
+/**
+ * Add a new model pool configuration
+ */
+export function addModelPool(
+  baseDir: string,
+  poolConfig: Omit<import('../types/litellm-api-config.js').ModelPoolConfig, 'id'>
+): { poolId: string; syncResult?: { success: boolean; message: string; endpointCount?: number } } {
+  const config = loadLiteLLMApiConfig(baseDir);
+  
+  // Auto-migrate if needed
+  migrateEmbeddingPoolToModelPools(config);
+  
+  // Ensure modelPools array exists
+  if (!config.modelPools) {
+    config.modelPools = [];
+  }
+
+  // Generate unique ID
+  const poolId = `pool-${poolConfig.modelType}-${Date.now()}`;
+  
+  const newPool: import('../types/litellm-api-config.js').ModelPoolConfig = {
+    ...poolConfig,
+    id: poolId,
+  };
+
+  config.modelPools.push(newPool);
+  saveConfig(baseDir, config);
+
+  // Sync to CodexLens if this is an embedding pool
+  const syncResult = poolConfig.modelType === 'embedding' && poolConfig.enabled
+    ? syncCodexLensConfig(baseDir)
+    : undefined;
+
+  return { poolId, syncResult };
+}
+
+/**
+ * Update an existing model pool configuration
+ */
+export function updateModelPool(
+  baseDir: string,
+  poolId: string,
+  updates: Partial<Omit<import('../types/litellm-api-config.js').ModelPoolConfig, 'id'>>
+): { success: boolean; syncResult?: { success: boolean; message: string; endpointCount?: number } } {
+  const config = loadLiteLLMApiConfig(baseDir);
+  
+  // Auto-migrate if needed
+  migrateEmbeddingPoolToModelPools(config);
+  
+  if (!config.modelPools) {
+    return { success: false };
+  }
+
+  const poolIndex = config.modelPools.findIndex(p => p.id === poolId);
+  if (poolIndex === -1) {
+    return { success: false };
+  }
+
+  // Apply updates
+  config.modelPools[poolIndex] = {
+    ...config.modelPools[poolIndex],
+    ...updates,
+  };
+
+  saveConfig(baseDir, config);
+
+  // Sync to CodexLens if this is an enabled embedding pool
+  const pool = config.modelPools[poolIndex];
+  const syncResult = pool.modelType === 'embedding' && pool.enabled
+    ? syncCodexLensConfig(baseDir)
+    : undefined;
+
+  return { success: true, syncResult };
+}
+
+/**
+ * Delete a model pool configuration
+ */
+export function deleteModelPool(
+  baseDir: string,
+  poolId: string
+): { success: boolean; syncResult?: { success: boolean; message: string; endpointCount?: number } } {
+  const config = loadLiteLLMApiConfig(baseDir);
+  
+  if (!config.modelPools) {
+    return { success: false };
+  }
+
+  const poolIndex = config.modelPools.findIndex(p => p.id === poolId);
+  if (poolIndex === -1) {
+    return { success: false };
+  }
+
+  const deletedPool = config.modelPools[poolIndex];
+  config.modelPools.splice(poolIndex, 1);
+  
+  saveConfig(baseDir, config);
+
+  // Sync to CodexLens if we deleted an embedding pool
+  const syncResult = deletedPool.modelType === 'embedding'
+    ? syncCodexLensConfig(baseDir)
+    : undefined;
+
+  return { success: true, syncResult };
+}
+
+/**
+ * Get available models for a specific model type
+ * Used for pool configuration UI
+ */
+export function getAvailableModelsForType(
+  baseDir: string,
+  modelType: import('../types/litellm-api-config.js').ModelPoolType
+): Array<{ modelId: string; modelName: string; providers: string[] }> {
+  const config = loadLiteLLMApiConfig(baseDir);
+  const availableModels: Array<{ modelId: string; modelName: string; providers: string[] }> = [];
+  const modelMap = new Map<string, { modelId: string; modelName: string; providers: string[] }>();
+
+  for (const provider of config.providers) {
+    if (!provider.enabled) continue;
+
+    let models: typeof provider.embeddingModels | undefined;
+    
+    switch (modelType) {
+      case 'embedding':
+        models = provider.embeddingModels;
+        break;
+      case 'llm':
+        models = provider.llmModels;
+        break;
+      case 'reranker':
+        models = provider.rerankerModels;
+        break;
+    }
+
+    if (!models) continue;
+
+    for (const model of models) {
+      if (!model.enabled) continue;
+
+      const key = model.id;
+      if (modelMap.has(key)) {
+        modelMap.get(key)!.providers.push(provider.name);
+      } else {
+        modelMap.set(key, {
+          modelId: model.id,
+          modelName: model.name,
+          providers: [provider.name],
+        });
+      }
+    }
+  }
+
+  availableModels.push(...Array.from(modelMap.values()));
+  return availableModels;
+}
+
 // Re-export types
 export type { ProviderCredential, CustomEndpoint, ProviderType, CacheStrategy, CodexLensEmbeddingRotation, CodexLensEmbeddingProvider, EmbeddingPoolConfig };
diff --git a/ccw/src/core/routes/cli-routes.ts b/ccw/src/core/routes/cli-routes.ts
index c6b4e618..3c75a599 100644
--- a/ccw/src/core/routes/cli-routes.ts
+++ b/ccw/src/core/routes/cli-routes.ts
@@ -34,6 +34,8 @@ import {
 import {
   loadClaudeCliTools,
   saveClaudeCliTools,
+  loadClaudeCliSettings,
+  saveClaudeCliSettings,
   updateClaudeToolEnabled,
   updateClaudeCacheSettings,
   getClaudeCliToolsInfo,
@@ -704,11 +706,13 @@ export async function handleCliRoutes(ctx: RouteContext): Promise<boolean> {
   // API: Get CLI Tools Config from .claude/cli-tools.json (with fallback to global)
   if (pathname === '/api/cli/tools-config' && req.method === 'GET') {
     try {
-      const config = loadClaudeCliTools(initialPath);
+      const toolsConfig = loadClaudeCliTools(initialPath);
+      const settingsConfig = loadClaudeCliSettings(initialPath);
       const info = getClaudeCliToolsInfo(initialPath);
       res.writeHead(200, { 'Content-Type': 'application/json' });
       res.end(JSON.stringify({
-        ...config,
+        tools: toolsConfig,
+        settings: settingsConfig,
         _configInfo: info
       }));
     } catch (err) {
@@ -722,32 +726,55 @@ export async function handleCliRoutes(ctx: RouteContext): Promise<boolean> {
   if (pathname === '/api/cli/tools-config' && req.method === 'PUT') {
     handlePostRequest(req, res, async (body: unknown) => {
       try {
-        const updates = body as Partial<any>;
-        const config = loadClaudeCliTools(initialPath);
+        const updates = body as { tools?: any; settings?: any };
 
-        // Merge updates
-        const updatedConfig = {
-          ...config,
-          ...updates,
-          tools: { ...config.tools, ...(updates.tools || {}) },
-          settings: {
-            ...config.settings,
-            ...(updates.settings || {}),
+        // Update tools config if provided
+        if (updates.tools) {
+          const currentTools = loadClaudeCliTools(initialPath);
+          const updatedTools = {
+            ...currentTools,
+            tools: { ...currentTools.tools, ...(updates.tools.tools || {}) },
+            customEndpoints: updates.tools.customEndpoints || currentTools.customEndpoints
+          };
+          saveClaudeCliTools(initialPath, updatedTools);
+        }
+
+        // Update settings config if provided
+        if (updates.settings) {
+          const currentSettings = loadClaudeCliSettings(initialPath);
+          const s = updates.settings;
+
+          // Deep merge: only update fields that are explicitly provided
+          const updatedSettings = {
+            ...currentSettings,
+            // Scalar fields: only update if explicitly provided
+            ...(s.defaultTool !== undefined && { defaultTool: s.defaultTool }),
+            ...(s.promptFormat !== undefined && { promptFormat: s.promptFormat }),
+            ...(s.nativeResume !== undefined && { nativeResume: s.nativeResume }),
+            ...(s.recursiveQuery !== undefined && { recursiveQuery: s.recursiveQuery }),
+            ...(s.codeIndexMcp !== undefined && { codeIndexMcp: s.codeIndexMcp }),
+            // Nested objects: deep merge
+            smartContext: {
+              ...currentSettings.smartContext,
+              ...(s.smartContext || {})
+            },
             cache: {
-              ...config.settings.cache,
-              ...(updates.settings?.cache || {})
+              ...currentSettings.cache,
+              ...(s.cache || {})
             }
-          }
-        };
+          };
+          saveClaudeCliSettings(initialPath, updatedSettings);
+        }
 
-        saveClaudeCliTools(initialPath, updatedConfig);
+        const toolsConfig = loadClaudeCliTools(initialPath);
+        const settingsConfig = loadClaudeCliSettings(initialPath);
 
         broadcastToClients({
           type: 'CLI_TOOLS_CONFIG_UPDATED',
-          payload: { config: updatedConfig, timestamp: new Date().toISOString() }
+          payload: { tools: toolsConfig, settings: settingsConfig, timestamp: new Date().toISOString() }
         });
 
-        return { success: true, config: updatedConfig };
+        return { success: true, tools: toolsConfig, settings: settingsConfig };
       } catch (err) {
         return { error: (err as Error).message, status: 500 };
       }
@@ -782,14 +809,14 @@ export async function handleCliRoutes(ctx: RouteContext): Promise<boolean> {
     handlePostRequest(req, res, async (body: unknown) => {
       try {
         const cacheSettings = body as { injectionMode?: string; defaultPrefix?: string; defaultSuffix?: string };
-        const config = updateClaudeCacheSettings(initialPath, cacheSettings as any);
+        const settings = updateClaudeCacheSettings(initialPath, cacheSettings as any);
 
         broadcastToClients({
           type: 'CLI_CACHE_SETTINGS_UPDATED',
-          payload: { cache: config.settings.cache, timestamp: new Date().toISOString() }
+          payload: { cache: settings.cache, timestamp: new Date().toISOString() }
         });
 
-        return { success: true, config };
+        return { success: true, settings };
       } catch (err) {
         return { error: (err as Error).message, status: 500 };
       }
diff --git a/ccw/src/core/routes/litellm-api-routes.ts b/ccw/src/core/routes/litellm-api-routes.ts
index 6624cf94..384f8f77 100644
--- a/ccw/src/core/routes/litellm-api-routes.ts
+++ b/ccw/src/core/routes/litellm-api-routes.ts
@@ -4,9 +4,35 @@
  */
 import { fileURLToPath } from 'url';
 import { dirname, join as pathJoin } from 'path';
+import { z } from 'zod';
 import { getSystemPython } from '../../utils/python-utils.js';
 import type { RouteContext } from './types.js';
 
+// ========== Input Validation Schemas ==========
+
+/**
+ * Validation schema for ModelPoolConfig
+ * Used to validate incoming API requests for model pool operations
+ */
+const ModelPoolConfigSchema = z.object({
+  modelType: z.enum(['embedding', 'llm', 'reranker']),
+  enabled: z.boolean(),
+  targetModel: z.string().min(1, 'Target model is required'),
+  strategy: z.enum(['round_robin', 'latency_aware', 'weighted_random']),
+  autoDiscover: z.boolean(),
+  excludedProviderIds: z.array(z.string()).optional().default([]),
+  defaultCooldown: z.number().int().min(0).default(60),
+  defaultMaxConcurrentPerKey: z.number().int().min(1).default(4),
+  name: z.string().optional(),
+  description: z.string().optional(),
+});
+
+/**
+ * Partial schema for updating ModelPoolConfig
+ * All fields are optional for PATCH-like updates
+ */
+const ModelPoolConfigUpdateSchema = ModelPoolConfigSchema.partial();
+
 // Get current module path for package-relative lookups
 const __filename = fileURLToPath(import.meta.url);
 const __dirname = dirname(__filename);
@@ -39,6 +65,12 @@ import {
   getEmbeddingPoolConfig,
   updateEmbeddingPoolConfig,
   discoverProvidersForModel,
+  getModelPools,
+  getModelPool,
+  addModelPool,
+  updateModelPool,
+  deleteModelPool,
+  getAvailableModelsForType,
   type ProviderCredential,
   type CustomEndpoint,
   type ProviderType,
@@ -856,6 +888,186 @@ export async function handleLiteLLMApiRoutes(ctx: RouteContext): Promise<boolean
     return true;
   }
 
+  // ========== Multi-Model Pool Management ==========
+
+  // GET /api/litellm-api/model-pools - Get all model pool configurations
+  if (pathname === '/api/litellm-api/model-pools' && req.method === 'GET') {
+    try {
+      const pools = getModelPools(initialPath);
+      
+      res.writeHead(200, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({ pools }));
+    } catch (err) {
+      res.writeHead(500, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({ error: (err as Error).message }));
+    }
+    return true;
+  }
+
+  // GET /api/litellm-api/model-pools/:id - Get specific pool configuration
+  const poolGetMatch = pathname.match(/^\/api\/litellm-api\/model-pools\/([^/]+)$/);
+  if (poolGetMatch && req.method === 'GET') {
+    const poolId = decodeURIComponent(poolGetMatch[1]);
+
+    try {
+      const pool = getModelPool(initialPath, poolId);
+      
+      if (!pool) {
+        res.writeHead(404, { 'Content-Type': 'application/json' });
+        res.end(JSON.stringify({ error: 'Pool not found' }));
+        return true;
+      }
+
+      res.writeHead(200, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({ pool }));
+    } catch (err) {
+      res.writeHead(500, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({ error: (err as Error).message }));
+    }
+    return true;
+  }
+
+  // POST /api/litellm-api/model-pools - Create new model pool
+  if (pathname === '/api/litellm-api/model-pools' && req.method === 'POST') {
+    handlePostRequest(req, res, async (body: unknown) => {
+      // Validate input using zod schema
+      const validationResult = ModelPoolConfigSchema.safeParse(body);
+      if (!validationResult.success) {
+        return {
+          error: 'Invalid request body',
+          details: validationResult.error.issues.map(e => ({
+            field: String(e.path.join('.')),
+            message: e.message
+          })),
+          status: 400
+        };
+      }
+
+      try {
+        const poolConfig = validationResult.data;
+        const result = addModelPool(initialPath, poolConfig);
+
+        broadcastToClients({
+          type: 'MODEL_POOL_CREATED',
+          payload: { poolId: result.poolId, timestamp: new Date().toISOString() }
+        });
+
+        return { success: true, ...result };
+      } catch (err) {
+        return { error: (err as Error).message, status: 500 };
+      }
+    });
+    return true;
+  }
+
+  // PUT /api/litellm-api/model-pools/:id - Update model pool
+  const poolPutMatch = pathname.match(/^\/api\/litellm-api\/model-pools\/([^/]+)$/);
+  if (poolPutMatch && req.method === 'PUT') {
+    const poolId = decodeURIComponent(poolPutMatch[1]);
+
+    handlePostRequest(req, res, async (body: unknown) => {
+      // Validate input using partial schema (all fields optional for updates)
+      const validationResult = ModelPoolConfigUpdateSchema.safeParse(body);
+      if (!validationResult.success) {
+        return {
+          error: 'Invalid request body',
+          details: validationResult.error.issues.map(e => ({
+            field: String(e.path.join('.')),
+            message: e.message
+          })),
+          status: 400
+        };
+      }
+
+      try {
+        const updates = validationResult.data;
+        const result = updateModelPool(initialPath, poolId, updates);
+
+        if (!result.success) {
+          return { error: 'Pool not found', status: 404 };
+        }
+
+        broadcastToClients({
+          type: 'MODEL_POOL_UPDATED',
+          payload: { poolId, syncResult: result.syncResult, timestamp: new Date().toISOString() }
+        });
+
+        return result;
+      } catch (err) {
+        return { error: (err as Error).message, status: 500 };
+      }
+    });
+    return true;
+  }
+
+  // DELETE /api/litellm-api/model-pools/:id - Delete model pool
+  const poolDeleteMatch = pathname.match(/^\/api\/litellm-api\/model-pools\/([^/]+)$/);
+  if (poolDeleteMatch && req.method === 'DELETE') {
+    const poolId = decodeURIComponent(poolDeleteMatch[1]);
+
+    try {
+      const result = deleteModelPool(initialPath, poolId);
+
+      if (!result.success) {
+        res.writeHead(404, { 'Content-Type': 'application/json' });
+        res.end(JSON.stringify({ error: 'Pool not found' }));
+        return true;
+      }
+
+      broadcastToClients({
+        type: 'MODEL_POOL_DELETED',
+        payload: { poolId, syncResult: result.syncResult, timestamp: new Date().toISOString() }
+      });
+
+      res.writeHead(200, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify(result));
+    } catch (err) {
+      res.writeHead(500, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({ error: (err as Error).message }));
+    }
+    return true;
+  }
+
+  // GET /api/litellm-api/model-pools/available-models/:modelType - Get available models for type
+  const availableModelsMatch = pathname.match(/^\/api\/litellm-api\/model-pools\/available-models\/([^/]+)$/);
+  if (availableModelsMatch && req.method === 'GET') {
+    const modelType = decodeURIComponent(availableModelsMatch[1]) as import('../../types/litellm-api-config.js').ModelPoolType;
+
+    try {
+      const availableModels = getAvailableModelsForType(initialPath, modelType);
+
+      res.writeHead(200, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({ availableModels, modelType }));
+    } catch (err) {
+      res.writeHead(500, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({ error: (err as Error).message }));
+    }
+    return true;
+  }
+
+  // GET /api/litellm-api/model-pools/discover/:modelType/:model - Discover providers for model
+  const discoverPoolMatch = pathname.match(/^\/api\/litellm-api\/model-pools\/discover\/([^/]+)\/([^/]+)$/);
+  if (discoverPoolMatch && req.method === 'GET') {
+    const modelType = decodeURIComponent(discoverPoolMatch[1]);
+    const targetModel = decodeURIComponent(discoverPoolMatch[2]);
+
+    try {
+      const discovered = discoverProvidersForModel(initialPath, targetModel);
+
+      res.writeHead(200, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({
+        modelType,
+        targetModel,
+        discovered,
+        count: discovered.length,
+      }));
+    } catch (err) {
+      res.writeHead(500, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({ error: (err as Error).message }));
+    }
+    return true;
+  }
+
   // POST /api/litellm-api/ccw-litellm/install - Install ccw-litellm package
   if (pathname === '/api/litellm-api/ccw-litellm/install' && req.method === 'POST') {
     handlePostRequest(req, res, async () => {
diff --git a/ccw/src/core/server.ts b/ccw/src/core/server.ts
index 6d8a8b8c..9b90f68a 100644
--- a/ccw/src/core/server.ts
+++ b/ccw/src/core/server.ts
@@ -550,15 +550,15 @@ export async function startServer(options: ServerOptions = {}): Promise<http.Ser
 
       // Serve dashboard HTML
       if (pathname === '/' || pathname === '/index.html') {
-        if (isLocalhostRequest(req)) {
-          const tokenResult = tokenManager.getOrCreateAuthToken();
-          setAuthCookie(res, tokenResult.token, tokenResult.expiresAt);
+        // Set session cookie and CSRF token for all requests
+        const tokenResult = tokenManager.getOrCreateAuthToken();
+        setAuthCookie(res, tokenResult.token, tokenResult.expiresAt);
+
+        const sessionId = getOrCreateSessionId(req, res);
+        const csrfToken = getCsrfTokenManager().generateToken(sessionId);
+        res.setHeader('X-CSRF-Token', csrfToken);
+        setCsrfCookie(res, csrfToken, 15 * 60);
 
-          const sessionId = getOrCreateSessionId(req, res);
-          const csrfToken = getCsrfTokenManager().generateToken(sessionId);
-          res.setHeader('X-CSRF-Token', csrfToken);
-          setCsrfCookie(res, csrfToken, 15 * 60);
-        }
         const html = generateServerDashboard(initialPath);
         res.writeHead(200, { 'Content-Type': 'text/html; charset=utf-8' });
         res.end(html);
diff --git a/ccw/src/templates/dashboard-css/31-api-settings.css b/ccw/src/templates/dashboard-css/31-api-settings.css
index 4e1685a6..90fce5ce 100644
--- a/ccw/src/templates/dashboard-css/31-api-settings.css
+++ b/ccw/src/templates/dashboard-css/31-api-settings.css
@@ -972,27 +972,31 @@ select.cli-input {
 
 .sidebar-tabs {
   display: flex;
-  gap: 0;
+  flex-wrap: nowrap;
+  gap: 0.125rem;
   padding: 0.5rem;
   background: hsl(var(--muted) / 0.3);
   border-bottom: 1px solid hsl(var(--border));
 }
 
 .sidebar-tab {
-  flex: 1;
+  flex: 1 1 0;
   display: flex;
+  flex-direction: column;
   align-items: center;
   justify-content: center;
-  gap: 0.25rem;
+  gap: 0.125rem;
   padding: 0.5rem 0.25rem;
   font-size: 0.75rem;
   font-weight: 500;
   color: hsl(var(--muted-foreground));
   background: transparent;
-  border: none;
-  border-radius: 0.375rem;
+  border: 1px solid transparent;
+  border-radius: 0.5rem;
   cursor: pointer;
   transition: all 0.2s;
+  white-space: nowrap;
+  min-width: 0;
 }
 
 .sidebar-tab:hover {
@@ -1002,12 +1006,14 @@ select.cli-input {
 
 .sidebar-tab.active {
   color: hsl(var(--primary));
-  background: hsl(var(--primary) / 0.1);
+  background: hsl(var(--card));
+  border-color: hsl(var(--border));
+  box-shadow: 0 1px 3px hsl(var(--foreground) / 0.05);
 }
 
 .sidebar-tab i {
-  width: 14px;
-  height: 14px;
+  width: 18px;
+  height: 18px;
 }
 
 /* Sidebar content areas */
@@ -1107,13 +1113,13 @@ select.cli-input {
 /* Responsive adjustments for tabs */
 @media (max-width: 768px) {
   .sidebar-tab {
-    padding: 0.5rem 0.5rem;
-    font-size: 0.7rem;
+    padding: 0.5rem 0.625rem;
+    font-size: 0.75rem;
   }
 
   .sidebar-tab i {
-    width: 12px;
-    height: 12px;
+    width: 16px;
+    height: 16px;
   }
 }
 
diff --git a/ccw/src/templates/dashboard-js/views/api-settings.js b/ccw/src/templates/dashboard-js/views/api-settings.js
index 10895641..aa49d275 100644
--- a/ccw/src/templates/dashboard-js/views/api-settings.js
+++ b/ccw/src/templates/dashboard-js/views/api-settings.js
@@ -11,13 +11,19 @@ let selectedProviderId = null;
 let providerSearchQuery = '';
 let activeModelTab = 'llm';
 let expandedModelGroups = new Set();
-let activeSidebarTab = 'providers'; // 'providers' | 'endpoints' | 'cache' | 'embedding-pool' | 'cli-settings'
+let activeSidebarTab = 'providers'; // 'providers' | 'endpoints' | 'cache' | 'embedding-pool' | 'model-pools' | 'cli-settings'
 
-// Embedding Pool state
+// Embedding Pool state (legacy, kept for backward compatibility)
 let embeddingPoolConfig = null;
 let embeddingPoolAvailableModels = [];
 let embeddingPoolDiscoveredProviders = [];
 
+// Multi-Model Pool state
+let modelPools = [];
+let selectedPoolId = null;
+let poolAvailableModels = {};
+let poolDiscoveredProviders = {};
+
 // CLI Settings state
 let cliSettingsData = null;
 let selectedCliSettingsId = null;
@@ -30,6 +36,9 @@ const CCW_LITELLM_STATUS_CACHE_TTL = 60000; // 60 seconds
 // Track if this is the first render (force refresh on first load)
 let isFirstApiSettingsRender = true;
 
+// Note: CSRF token management (csrfToken, initCsrfToken, csrfFetch) is defined in cli-manager.js
+// and shared across all views when files are bundled together
+
 // ========== Data Loading ==========
 
 /**
@@ -130,6 +139,57 @@ async function loadCliSettings(forceRefresh = false) {
   }
 }
 
+/**
+ * Load all model pool configurations
+ */
+async function loadModelPools() {
+  try {
+    const response = await fetch('/api/litellm-api/model-pools');
+    if (!response.ok) throw new Error('Failed to load model pools');
+    const data = await response.json();
+    modelPools = data.pools || [];
+    return modelPools;
+  } catch (err) {
+    console.error('Failed to load model pools:', err);
+    showRefreshToast(t('common.error') + ': ' + err.message, 'error');
+    return [];
+  }
+}
+
+/**
+ * Load available models for a specific model type
+ */
+async function loadAvailableModelsForType(modelType) {
+  try {
+    const response = await fetch('/api/litellm-api/model-pools/available-models/' + modelType);
+    if (!response.ok) throw new Error('Failed to load available models');
+    const data = await response.json();
+    poolAvailableModels[modelType] = data.availableModels || [];
+    return data.availableModels;
+  } catch (err) {
+    console.error('Failed to load available models:', err);
+    return [];
+  }
+}
+
+/**
+ * Discover providers for a specific model in pool context
+ */
+async function discoverProvidersForPool(modelType, targetModel) {
+  try {
+    const response = await fetch('/api/litellm-api/model-pools/discover/' + modelType + '/' + encodeURIComponent(targetModel));
+    if (!response.ok) throw new Error('Failed to discover providers');
+    const data = await response.json();
+    const key = modelType + ':' + targetModel;
+    poolDiscoveredProviders[key] = data.discovered || [];
+    return data;
+  } catch (err) {
+    console.error('Failed to discover providers:', err);
+    poolDiscoveredProviders[key] = [];
+    return null;
+  }
+}
+
 /**
  * Save CLI Settings endpoint
  */
@@ -138,7 +198,8 @@ async function saveCliSettingsEndpoint(data) {
     const method = data.id ? 'PUT' : 'POST';
     const url = data.id ? '/api/cli/settings/' + data.id : '/api/cli/settings';
 
-    const response = await fetch(url, {
+    await initCsrfToken();
+    const response = await csrfFetch(url, {
       method: method,
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify(data)
@@ -174,7 +235,8 @@ async function deleteCliSettingsEndpoint(endpointId) {
   if (!confirm(t('apiSettings.confirmDeleteSettings'))) return;
 
   try {
-    const response = await fetch('/api/cli/settings/' + endpointId, {
+    await initCsrfToken();
+    const response = await csrfFetch('/api/cli/settings/' + endpointId, {
       method: 'DELETE'
     });
 
@@ -237,7 +299,8 @@ async function saveEmbeddingPoolConfig() {
       defaultMaxConcurrentPerKey: defaultMaxConcurrentPerKey
     } : null;
 
-    const response = await fetch('/api/litellm-api/embedding-pool', {
+    await initCsrfToken();
+    const response = await csrfFetch('/api/litellm-api/embedding-pool', {
       method: 'PUT',
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify(poolConfig)
@@ -260,9 +323,9 @@ async function saveEmbeddingPoolConfig() {
     // Update sidebar summary
     const sidebarContainer = document.querySelector('.api-settings-sidebar');
     if (sidebarContainer) {
-      const contentArea = sidebarContainer.querySelector('.provider-list, .endpoints-list, .embedding-pool-sidebar-info, .embedding-pool-sidebar-summary, .cache-sidebar-info');
-      if (contentArea && contentArea.parentElement) {
-        contentArea.parentElement.innerHTML = renderEmbeddingPoolSidebar();
+      const contentArea = sidebarContainer.querySelector('.embedding-pool-sidebar-info, .embedding-pool-sidebar-summary');
+      if (contentArea) {
+        contentArea.outerHTML = renderEmbeddingPoolSidebar();
         if (window.lucide) lucide.createIcons();
       }
     }
@@ -292,11 +355,11 @@ async function toggleProviderExclusion(providerId) {
 
   // Re-render the discovered providers section
   renderDiscoveredProviders();
-  
+
   // Update sidebar summary
-  const sidebarContainer = document.querySelector('.api-settings-sidebar .embedding-pool-sidebar-summary');
-  if (sidebarContainer && sidebarContainer.parentElement) {
-    sidebarContainer.parentElement.innerHTML = renderEmbeddingPoolSidebar();
+  const sidebarContainer = document.querySelector('.api-settings-sidebar .embedding-pool-sidebar-summary, .api-settings-sidebar .embedding-pool-sidebar-info');
+  if (sidebarContainer) {
+    sidebarContainer.outerHTML = renderEmbeddingPoolSidebar();
     if (window.lucide) lucide.createIcons();
   }
 }
@@ -557,7 +620,8 @@ async function saveProvider() {
       : '/api/litellm-api/providers';
     const method = providerId ? 'PUT' : 'POST';
 
-    const response = await fetch(url, {
+    await initCsrfToken();
+    const response = await csrfFetch(url, {
       method: method,
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify(providerData)
@@ -585,7 +649,8 @@ async function deleteProvider(providerId) {
   if (!confirm(t('apiSettings.confirmDeleteProvider'))) return;
 
   try {
-    const response = await fetch('/api/litellm-api/providers/' + providerId, {
+    await initCsrfToken();
+    const response = await csrfFetch('/api/litellm-api/providers/' + providerId, {
       method: 'DELETE'
     });
 
@@ -624,7 +689,8 @@ async function testProviderConnection(providerIdParam) {
   }
 
   try {
-    const response = await fetch('/api/litellm-api/providers/' + providerId + '/test', {
+    await initCsrfToken();
+    const response = await csrfFetch('/api/litellm-api/providers/' + providerId + '/test', {
       method: 'POST'
     });
 
@@ -917,7 +983,8 @@ async function saveEndpoint() {
       : '/api/litellm-api/endpoints';
     const method = form.dataset.endpointId ? 'PUT' : 'POST';
 
-    const response = await fetch(url, {
+    await initCsrfToken();
+    const response = await csrfFetch(url, {
       method: method,
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify(endpointData)
@@ -945,7 +1012,8 @@ async function deleteEndpoint(endpointId) {
   if (!confirm(t('apiSettings.confirmDeleteEndpoint'))) return;
 
   try {
-    const response = await fetch('/api/litellm-api/endpoints/' + endpointId, {
+    await initCsrfToken();
+    const response = await csrfFetch('/api/litellm-api/endpoints/' + endpointId, {
       method: 'DELETE'
     });
 
@@ -1018,7 +1086,8 @@ async function clearCache() {
   if (!confirm(t('apiSettings.confirmClearCache'))) return;
 
   try {
-    const response = await fetch('/api/litellm-api/cache/clear', {
+    await initCsrfToken();
+    const response = await csrfFetch('/api/litellm-api/cache/clear', {
       method: 'POST'
     });
 
@@ -1042,7 +1111,8 @@ async function toggleGlobalCache() {
   const enabled = document.getElementById('global-cache-enabled').checked;
 
   try {
-    const response = await fetch('/api/litellm-api/config/cache', {
+    await initCsrfToken();
+    const response = await csrfFetch('/api/litellm-api/config/cache', {
       method: 'PUT',
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify({ enabled: enabled })
@@ -1095,8 +1165,11 @@ async function renderApiSettings() {
     '<button class="sidebar-tab' + (activeSidebarTab === 'cli-settings' ? ' active' : '') + '" onclick="switchSidebarTab(\'cli-settings\')">' +
     '<i data-lucide="settings"></i> ' + t('apiSettings.cliSettings') +
     '</button>' +
+    '<button class="sidebar-tab' + (activeSidebarTab === 'model-pools' ? ' active' : '') + '" onclick="switchSidebarTab(\'model-pools\')">' +
+    '<i data-lucide="layers"></i> ' + t('apiSettings.modelPools') +
+    '</button>' +
     '<button class="sidebar-tab' + (activeSidebarTab === 'embedding-pool' ? ' active' : '') + '" onclick="switchSidebarTab(\'embedding-pool\')">' +
-    '<i data-lucide="repeat"></i> ' + t('apiSettings.embeddingPool') +
+    '<i data-lucide="repeat"></i> ' + t('apiSettings.embeddingPool') + ' (Legacy)' +
     '</button>' +
     '<button class="sidebar-tab' + (activeSidebarTab === 'cache' ? ' active' : '') + '" onclick="switchSidebarTab(\'cache\')">' +
     '<i data-lucide="database"></i> ' + t('apiSettings.cache') +
@@ -1127,6 +1200,15 @@ async function renderApiSettings() {
       await loadEmbeddingPoolConfig();
     }
     sidebarContentHtml = renderEmbeddingPoolSidebar();
+  } else if (activeSidebarTab === 'model-pools') {
+    // Load model pools first if not already loaded
+    if (!modelPools || modelPools.length === 0) {
+      await loadModelPools();
+    }
+    sidebarContentHtml = '<div class="model-pools-list" id="model-pools-list"></div>';
+    addButtonHtml = '<button class="btn btn-primary btn-full" onclick="showAddModelPoolModal()">' +
+      '<i data-lucide="plus"></i> ' + t('apiSettings.addModelPool') +
+      '</button>';
   } else if (activeSidebarTab === 'cache') {
     sidebarContentHtml = '<div class="cache-sidebar-info" style="padding: 1rem; color: var(--text-secondary); font-size: 0.875rem;">' +
       '<p>' + t('apiSettings.cacheTabHint') + '</p>' +
@@ -1177,6 +1259,16 @@ async function renderApiSettings() {
     renderEndpointsMainPanel();
   } else if (activeSidebarTab === 'embedding-pool') {
     renderEmbeddingPoolMainPanel();
+  } else if (activeSidebarTab === 'model-pools') {
+    renderModelPoolsList();
+    // Auto-select first pool if exists
+    if (!selectedPoolId && modelPools && modelPools.length > 0) {
+      selectModelPool(modelPools[0].id);
+    } else if (selectedPoolId) {
+      renderModelPoolDetail(selectedPoolId);
+    } else {
+      renderModelPoolEmptyState();
+    }
   } else if (activeSidebarTab === 'cache') {
     renderCacheMainPanel();
   } else if (activeSidebarTab === 'cli-settings') {
@@ -1574,7 +1666,8 @@ function getDefaultApiBase(type) {
  */
 async function toggleProviderEnabled(providerId, enabled) {
   try {
-    var response = await fetch('/api/litellm-api/providers/' + providerId, {
+    await initCsrfToken();
+    var response = await csrfFetch('/api/litellm-api/providers/' + providerId, {
       method: 'PUT',
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify({ enabled: enabled })
@@ -2015,7 +2108,7 @@ function saveNewModel(event, providerId, modelType) {
       }
 
       models.push(newModel);
-      return fetch('/api/litellm-api/providers/' + providerId, {
+      return csrfFetch('/api/litellm-api/providers/' + providerId, {
         method: 'PUT',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify({ [modelsKey]: models })
@@ -2260,7 +2353,7 @@ function saveModelSettings(event, providerId, modelId, modelType) {
       var updateData = {};
       updateData[modelsKey] = models;
 
-      return fetch('/api/litellm-api/providers/' + providerId, {
+      return csrfFetch('/api/litellm-api/providers/' + providerId, {
         method: 'PUT',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify(updateData)
@@ -2299,7 +2392,7 @@ function deleteModel(providerId, modelId, modelType) {
       var updateData = {};
       updateData[modelsKey] = updatedModels;
 
-      return fetch('/api/litellm-api/providers/' + providerId, {
+      return csrfFetch('/api/litellm-api/providers/' + providerId, {
         method: 'PUT',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify(updateData)
@@ -2342,7 +2435,8 @@ async function saveProviderApiBase(providerId) {
   }
 
   try {
-    var response = await fetch('/api/litellm-api/providers/' + providerId, {
+    await initCsrfToken();
+    var response = await csrfFetch('/api/litellm-api/providers/' + providerId, {
       method: 'PUT',
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify({ apiBase: newApiBase || undefined })
@@ -2394,7 +2488,8 @@ async function deleteProviderWithConfirm(providerId) {
   if (!confirm(t('apiSettings.confirmDeleteProvider'))) return;
 
   try {
-    var response = await fetch('/api/litellm-api/providers/' + providerId, {
+    await initCsrfToken();
+    var response = await csrfFetch('/api/litellm-api/providers/' + providerId, {
       method: 'DELETE'
     });
 
@@ -2428,7 +2523,8 @@ async function deleteProviderWithConfirm(providerId) {
  */
 async function syncConfigToCodexLens() {
   try {
-    var response = await fetch('/api/litellm-api/config/sync', {
+    await initCsrfToken();
+    var response = await csrfFetch('/api/litellm-api/config/sync', {
       method: 'POST'
     });
 
@@ -2992,9 +3088,10 @@ async function onTargetModelChange(modelId) {
   // Update sidebar summary
   const sidebarContainer = document.querySelector('.api-settings-sidebar');
   if (sidebarContainer) {
-    const contentArea = sidebarContainer.querySelector('.provider-list, .endpoints-list, .embedding-pool-sidebar-info, .embedding-pool-sidebar-summary, .cache-sidebar-info');
-    if (contentArea && contentArea.parentElement) {
-      contentArea.parentElement.innerHTML = renderEmbeddingPoolSidebar();
+    const contentArea = sidebarContainer.querySelector('.provider-list, .endpoints-list, .embedding-pool-sidebar-info, .embedding-pool-sidebar-summary, .cache-sidebar-info, .cli-settings-list');
+    if (contentArea) {
+      // Use outerHTML to replace only the content area, not the entire sidebar
+      contentArea.outerHTML = renderEmbeddingPoolSidebar();
       if (window.lucide) lucide.createIcons();
     }
   }
@@ -3253,7 +3350,7 @@ function addApiKey(providerId) {
     .then(function(provider) {
       const apiKeys = provider.apiKeys || [];
       apiKeys.push(newKey);
-      return fetch('/api/litellm-api/providers/' + providerId, {
+      return csrfFetch('/api/litellm-api/providers/' + providerId, {
         method: 'PUT',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify({ apiKeys: apiKeys })
@@ -3279,7 +3376,7 @@ function removeApiKey(providerId, keyId) {
     .then(function(res) { return res.json(); })
     .then(function(provider) {
       const apiKeys = (provider.apiKeys || []).filter(function(k) { return k.id !== keyId; });
-      return fetch('/api/litellm-api/providers/' + providerId, {
+      return csrfFetch('/api/litellm-api/providers/' + providerId, {
         method: 'PUT',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify({ apiKeys: apiKeys })
@@ -3307,7 +3404,7 @@ function updateApiKeyField(providerId, keyId, field, value) {
       if (keyIndex >= 0) {
         apiKeys[keyIndex][field] = value;
       }
-      return fetch('/api/litellm-api/providers/' + providerId, {
+      return csrfFetch('/api/litellm-api/providers/' + providerId, {
         method: 'PUT',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify({ apiKeys: apiKeys })
@@ -3322,7 +3419,7 @@ function updateApiKeyField(providerId, keyId, field, value) {
  * Update provider routing strategy
  */
 function updateProviderRouting(providerId, strategy) {
-  fetch('/api/litellm-api/providers/' + providerId, {
+  csrfFetch('/api/litellm-api/providers/' + providerId, {
     method: 'PUT',
     headers: { 'Content-Type': 'application/json' },
     body: JSON.stringify({ routingStrategy: strategy })
@@ -3340,7 +3437,7 @@ function updateHealthCheckEnabled(providerId, enabled) {
     .then(function(provider) {
       const healthCheck = provider.healthCheck || { intervalSeconds: 300, cooldownSeconds: 5, failureThreshold: 3 };
       healthCheck.enabled = enabled;
-      return fetch('/api/litellm-api/providers/' + providerId, {
+      return csrfFetch('/api/litellm-api/providers/' + providerId, {
         method: 'PUT',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify({ healthCheck: healthCheck })
@@ -3365,7 +3462,7 @@ function updateHealthCheckField(providerId, field, value) {
     .then(function(provider) {
       const healthCheck = provider.healthCheck || { enabled: false, intervalSeconds: 300, cooldownSeconds: 5, failureThreshold: 3 };
       healthCheck[field] = value;
-      return fetch('/api/litellm-api/providers/' + providerId, {
+      return csrfFetch('/api/litellm-api/providers/' + providerId, {
         method: 'PUT',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify({ healthCheck: healthCheck })
@@ -3385,7 +3482,7 @@ function testApiKey(providerId, keyId) {
   btn.classList.add('testing');
   btn.textContent = t('apiSettings.testingKey');
 
-  fetch('/api/litellm-api/providers/' + providerId + '/test-key', {
+  csrfFetch('/api/litellm-api/providers/' + providerId + '/test-key', {
     method: 'POST',
     headers: { 'Content-Type': 'application/json' },
     body: JSON.stringify({ keyId: keyId })
@@ -3525,7 +3622,8 @@ async function installCcwLitellm() {
   }
 
   try {
-    var response = await fetch('/api/litellm-api/ccw-litellm/install', {
+    await initCsrfToken();
+    var response = await csrfFetch('/api/litellm-api/ccw-litellm/install', {
       method: 'POST',
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify({})
@@ -3566,7 +3664,8 @@ async function uninstallCcwLitellm() {
   }
 
   try {
-    var response = await fetch('/api/litellm-api/ccw-litellm/uninstall', {
+    await initCsrfToken();
+    var response = await csrfFetch('/api/litellm-api/ccw-litellm/uninstall', {
       method: 'POST',
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify({})
@@ -3953,6 +4052,387 @@ async function submitCliSettings() {
   }
 }
 
+// ========== Multi-Model Pool Management ==========
+
+/**
+ * Render model pools list in sidebar
+ */
+function renderModelPoolsList() {
+  var container = document.getElementById('model-pools-list');
+  if (!container) return;
+
+  if (!modelPools || modelPools.length === 0) {
+    container.innerHTML = '<div class="empty-state" style="padding: 2rem; text-align: center; color: var(--text-secondary);">' +
+      '<i data-lucide="layers" style="width: 48px; height: 48px; margin-bottom: 1rem;"></i>' +
+      '<p>' + t('apiSettings.noPoolsConfigured') + '</p>' +
+      '</div>';
+    if (window.lucide) lucide.createIcons();
+    return;
+  }
+
+  // Group pools by type
+  var poolsByType = {
+    embedding: [],
+    llm: [],
+    reranker: []
+  };
+
+  modelPools.forEach(function(pool) {
+    if (poolsByType[pool.modelType]) {
+      poolsByType[pool.modelType].push(pool);
+    }
+  });
+
+  var html = '';
+  
+  // Render each type group
+  ['embedding', 'llm', 'reranker'].forEach(function(type) {
+    var pools = poolsByType[type];
+    if (pools.length === 0) return;
+
+    var typeLabel = type === 'embedding' ? t('apiSettings.embeddingPools') : 
+                    type === 'llm' ? t('apiSettings.llmPools') : 
+                    t('apiSettings.rerankerPools');
+
+    html += '<div class="pool-type-group" style="margin-bottom: 1.5rem;">' +
+      '<div class="pool-type-header" style="padding: 0.5rem; font-size: 0.75rem; font-weight: 600; text-transform: uppercase; color: var(--text-secondary); border-bottom: 1px solid var(--border);">' +
+      typeLabel +
+      '</div>';
+
+    pools.forEach(function(pool) {
+      var isSelected = selectedPoolId === pool.id;
+      var statusClass = pool.enabled ? 'status-enabled' : 'status-disabled';
+      var statusText = pool.enabled ? t('common.enabled') : t('common.disabled');
+
+      html += '<div class="pool-item' + (isSelected ? ' selected' : '') + '" onclick="selectModelPool(\'' + pool.id + '\')" style="padding: 0.75rem; cursor: pointer; border-bottom: 1px solid var(--border);">' +
+        '<div style="display: flex; justify-content: space-between; align-items: center;">' +
+        '<div style="flex: 1; min-width: 0;">' +
+        '<div style="font-weight: 500; margin-bottom: 0.25rem;">' + escapeHtml(pool.name || pool.targetModel) + '</div>' +
+        '<div style="font-size: 0.75rem; color: var(--text-secondary);">' + escapeHtml(pool.targetModel) + '</div>' +
+        '</div>' +
+        '<span class="status-badge ' + statusClass + '" style="font-size: 0.7rem; padding: 0.25rem 0.5rem; border-radius: 4px;">' + statusText + '</span>' +
+        '</div>' +
+        '</div>';
+    });
+
+    html += '</div>';
+  });
+
+  container.innerHTML = html;
+  if (window.lucide) lucide.createIcons();
+}
+
+/**
+ * Select a model pool
+ */
+function selectModelPool(poolId) {
+  selectedPoolId = poolId;
+  renderModelPoolsList();
+  renderModelPoolDetail(poolId);
+}
+
+/**
+ * Render model pool detail in main panel
+ */
+function renderModelPoolDetail(poolId) {
+  var container = document.getElementById('provider-detail-panel');
+  if (!container) return;
+
+  var pool = modelPools.find(function(p) { return p.id === poolId; });
+  if (!pool) {
+    renderModelPoolEmptyState();
+    return;
+  }
+
+  var typeLabel = pool.modelType === 'embedding' ? t('apiSettings.embedding') :
+                  pool.modelType === 'llm' ? t('apiSettings.llm') :
+                  t('apiSettings.reranker');
+
+  var html = '<div class="provider-detail">' +
+    '<div class="provider-detail-header">' +
+    '<div>' +
+    '<h2>' + escapeHtml(pool.name || pool.targetModel) + '</h2>' +
+    '<p style="color: var(--text-secondary); margin-top: 0.5rem;">' + typeLabel + ' Pool</p>' +
+    '</div>' +
+    '<div class="provider-actions">' +
+    '<button class="btn btn-secondary" onclick="editModelPool(\'' + pool.id + '\')"><i data-lucide="edit-2"></i> ' + t('common.edit') + '</button>' +
+    '<button class="btn btn-danger" onclick="deleteModelPool(\'' + pool.id + '\')"><i data-lucide="trash-2"></i> ' + t('common.delete') + '</button>' +
+    '</div>' +
+    '</div>' +
+    '<div class="provider-detail-body">' +
+    
+    // Basic Info
+    '<div class="form-section">' +
+    '<h3>' + t('apiSettings.basicInfo') + '</h3>' +
+    '<div class="info-grid">' +
+    '<div class="info-item"><label>' + t('apiSettings.status') + '</label><span class="status-badge ' + (pool.enabled ? 'status-enabled' : 'status-disabled') + '">' + (pool.enabled ? t('common.enabled') : t('common.disabled')) + '</span></div>' +
+    '<div class="info-item"><label>' + t('apiSettings.modelType') + '</label><span>' + typeLabel + '</span></div>' +
+    '<div class="info-item"><label>' + t('apiSettings.targetModel') + '</label><span>' + escapeHtml(pool.targetModel) + '</span></div>' +
+    '<div class="info-item"><label>' + t('apiSettings.strategy') + '</label><span>' + pool.strategy + '</span></div>' +
+    '<div class="info-item"><label>' + t('apiSettings.autoDiscover') + '</label><span>' + (pool.autoDiscover ? t('common.yes') : t('common.no')) + '</span></div>' +
+    '<div class="info-item"><label>' + t('apiSettings.cooldown') + '</label><span>' + pool.defaultCooldown + 's</span></div>' +
+    '<div class="info-item"><label>' + t('apiSettings.maxConcurrent') + '</label><span>' + pool.defaultMaxConcurrentPerKey + '</span></div>' +
+    '</div>' +
+    '</div>';
+
+  if (pool.description) {
+    html += '<div class="form-section">' +
+      '<h3>' + t('apiSettings.description') + '</h3>' +
+      '<p>' + escapeHtml(pool.description) + '</p>' +
+      '</div>';
+  }
+
+  // Excluded Providers
+  if (pool.excludedProviderIds && pool.excludedProviderIds.length > 0) {
+    html += '<div class="form-section">' +
+      '<h3>' + t('apiSettings.excludedProviders') + '</h3>' +
+      '<div class="excluded-providers-list">';
+    
+    pool.excludedProviderIds.forEach(function(providerId) {
+      html += '<span class="tag">' + escapeHtml(providerId) + '</span>';
+    });
+    
+    html += '</div></div>';
+  }
+
+  html += '</div></div>';
+
+  container.innerHTML = html;
+  if (window.lucide) lucide.createIcons();
+}
+
+/**
+ * Render empty state for model pools
+ */
+function renderModelPoolEmptyState() {
+  var container = document.getElementById('provider-detail-panel');
+  if (!container) return;
+
+  container.innerHTML = '<div class="empty-state">' +
+    '<i data-lucide="layers" style="width: 64px; height: 64px; margin-bottom: 1rem;"></i>' +
+    '<h3>' + t('apiSettings.noPoolSelected') + '</h3>' +
+    '<p>' + t('apiSettings.selectPoolFromList') + '</p>' +
+    '</div>';
+  
+  if (window.lucide) lucide.createIcons();
+}
+
+/**
+ * Show add model pool modal
+ */
+function showAddModelPoolModal() {
+  var modalHtml = '<div class="generic-modal-overlay active" id="add-pool-modal">' +
+    '<div class="generic-modal" style="max-width: 600px;">' +
+    '<div class="generic-modal-header">' +
+    '<h3 class="generic-modal-title">' + t('apiSettings.addModelPool') + '</h3>' +
+    '<button class="generic-modal-close" onclick="closeAddPoolModal()">&times;</button>' +
+    '</div>' +
+    '<div class="generic-modal-body">' +
+    '<form id="add-pool-form" class="api-settings-form" onsubmit="submitModelPool(event)">' +
+    
+    '<div class="form-group">' +
+    '<label>' + t('apiSettings.modelType') + ' *</label>' +
+    '<select id="pool-model-type" class="cli-input" required onchange="onPoolModelTypeChange()">' +
+    '<option value="">Select Type</option>' +
+    '<option value="embedding">Embedding</option>' +
+    '<option value="llm">LLM</option>' +
+    '<option value="reranker">Reranker</option>' +
+    '</select>' +
+    '</div>' +
+    
+    '<div class="form-group">' +
+    '<label>' + t('apiSettings.poolName') + '</label>' +
+    '<input type="text" id="pool-name" class="cli-input" placeholder="e.g., Primary Embedding Pool" />' +
+    '</div>' +
+    
+    '<div class="form-group">' +
+    '<label>' + t('apiSettings.targetModel') + ' *</label>' +
+    '<select id="pool-target-model" class="cli-input" required disabled>' +
+    '<option value="">Select model type first</option>' +
+    '</select>' +
+    '</div>' +
+    
+    '<div class="form-group">' +
+    '<label>' + t('apiSettings.strategy') + ' *</label>' +
+    '<select id="pool-strategy" class="cli-input" required>' +
+    '<option value="round_robin">Round Robin</option>' +
+    '<option value="latency_aware" selected>Latency Aware</option>' +
+    '<option value="weighted_random">Weighted Random</option>' +
+    '</select>' +
+    '</div>' +
+    
+    '<div class="form-group">' +
+    '<label>' + t('apiSettings.cooldown') + ' (seconds)</label>' +
+    '<input type="number" id="pool-cooldown" class="cli-input" value="60" min="0" />' +
+    '</div>' +
+    
+    '<div class="form-group">' +
+    '<label>' + t('apiSettings.maxConcurrent') + '</label>' +
+    '<input type="number" id="pool-max-concurrent" class="cli-input" value="4" min="1" />' +
+    '</div>' +
+    
+    '<div class="form-group">' +
+    '<label>' + t('apiSettings.description') + '</label>' +
+    '<textarea id="pool-description" class="cli-input" rows="2" placeholder="Optional description"></textarea>' +
+    '</div>' +
+    
+    '<div class="form-group">' +
+    '<label class="checkbox-label">' +
+    '<input type="checkbox" id="pool-enabled" checked /> ' + t('apiSettings.enablePool') +
+    '</label>' +
+    '</div>' +
+    
+    '<div class="form-group">' +
+    '<label class="checkbox-label">' +
+    '<input type="checkbox" id="pool-auto-discover" checked /> ' + t('apiSettings.autoDiscoverProviders') +
+    '</label>' +
+    '</div>' +
+    
+    '<div class="modal-actions">' +
+    '<button type="button" class="btn btn-secondary" onclick="closeAddPoolModal()"><i data-lucide="x"></i> ' + t('common.cancel') + '</button>' +
+    '<button type="submit" class="btn btn-primary"><i data-lucide="check"></i> ' + t('common.save') + '</button>' +
+    '</div>' +
+    '</form>' +
+    '</div>' +
+    '</div>' +
+    '</div>';
+
+  document.body.insertAdjacentHTML('beforeend', modalHtml);
+  if (window.lucide) lucide.createIcons();
+}
+
+/**
+ * Close add pool modal
+ */
+function closeAddPoolModal() {
+  var modal = document.getElementById('add-pool-modal');
+  if (modal) modal.remove();
+}
+
+/**
+ * Handle pool model type change
+ */
+async function onPoolModelTypeChange() {
+  var modelType = document.getElementById('pool-model-type').value;
+  var targetModelSelect = document.getElementById('pool-target-model');
+  
+  if (!modelType) {
+    targetModelSelect.disabled = true;
+    targetModelSelect.innerHTML = '<option value="">Select model type first</option>';
+    return;
+  }
+
+  // Load available models for this type
+  var models = await loadAvailableModelsForType(modelType);
+  
+  targetModelSelect.disabled = false;
+  targetModelSelect.innerHTML = '<option value="">Select a model</option>';
+  
+  models.forEach(function(model) {
+    var option = document.createElement('option');
+    option.value = model.modelId;
+    option.textContent = model.modelName + ' (' + model.providers.length + ' providers)';
+    targetModelSelect.appendChild(option);
+  });
+}
+
+/**
+ * Submit model pool form
+ */
+async function submitModelPool(event) {
+  event.preventDefault();
+
+  var poolData = {
+    modelType: document.getElementById('pool-model-type').value,
+    name: document.getElementById('pool-name').value,
+    targetModel: document.getElementById('pool-target-model').value,
+    strategy: document.getElementById('pool-strategy').value,
+    defaultCooldown: parseInt(document.getElementById('pool-cooldown').value),
+    defaultMaxConcurrentPerKey: parseInt(document.getElementById('pool-max-concurrent').value),
+    description: document.getElementById('pool-description').value,
+    enabled: document.getElementById('pool-enabled').checked,
+    autoDiscover: document.getElementById('pool-auto-discover').checked,
+    excludedProviderIds: []
+  };
+
+  try {
+    await initCsrfToken();
+    var response = await csrfFetch('/api/litellm-api/model-pools', {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify(poolData)
+    });
+
+    if (!response.ok) {
+      var err = await response.json();
+      throw new Error(err.error || 'Failed to create pool');
+    }
+
+    var result = await response.json();
+    showRefreshToast(t('apiSettings.poolCreated'), 'success');
+    
+    closeAddPoolModal();
+    
+    // Reload pools and switch to model-pools tab
+    await loadModelPools();
+    activeSidebarTab = 'model-pools';
+    renderApiSettings();
+  } catch (err) {
+    showRefreshToast(t('common.error') + ': ' + err.message, 'error');
+  }
+}
+
+/**
+ * Edit model pool
+ */
+function editModelPool(poolId) {
+  // TODO: Implement edit modal
+  showRefreshToast('Edit functionality coming soon', 'info');
+}
+
+/**
+ * Delete model pool
+ */
+async function deleteModelPool(poolId) {
+  if (!confirm(t('apiSettings.confirmDeletePool'))) {
+    return;
+  }
+
+  try {
+    await initCsrfToken();
+    var response = await csrfFetch('/api/litellm-api/model-pools/' + poolId, {
+      method: 'DELETE'
+    });
+
+    if (!response.ok) {
+      var err = await response.json();
+      throw new Error(err.error || 'Failed to delete pool');
+    }
+
+    showRefreshToast(t('apiSettings.poolDeleted'), 'success');
+    
+    // Reload pools
+    selectedPoolId = null;
+    await loadModelPools();
+    renderApiSettings();
+  } catch (err) {
+    showRefreshToast(t('common.error') + ': ' + err.message, 'error');
+  }
+}
+
+// Make model pool functions globally accessible
+window.loadModelPools = loadModelPools;
+window.renderModelPoolsList = renderModelPoolsList;
+window.selectModelPool = selectModelPool;
+window.renderModelPoolDetail = renderModelPoolDetail;
+window.renderModelPoolEmptyState = renderModelPoolEmptyState;
+window.showAddModelPoolModal = showAddModelPoolModal;
+window.closeAddPoolModal = closeAddPoolModal;
+window.onPoolModelTypeChange = onPoolModelTypeChange;
+window.submitModelPool = submitModelPool;
+window.editModelPool = editModelPool;
+window.deleteModelPool = deleteModelPool;
+
 // Make CLI Settings functions globally accessible
 window.loadCliSettings = loadCliSettings;
 window.saveCliSettingsEndpoint = saveCliSettingsEndpoint;
diff --git a/ccw/src/templates/dashboard-js/views/cli-manager.js b/ccw/src/templates/dashboard-js/views/cli-manager.js
index 3e6f3e84..9dea6a18 100644
--- a/ccw/src/templates/dashboard-js/views/cli-manager.js
+++ b/ccw/src/templates/dashboard-js/views/cli-manager.js
@@ -22,6 +22,12 @@ async function csrfFetch(url, options) {
 
   // Add CSRF token header for state-changing methods
   var method = (options.method || 'GET').toUpperCase();
+
+  // Auto-initialize CSRF token for state-changing requests
+  if (['POST', 'PUT', 'PATCH', 'DELETE'].indexOf(method) !== -1) {
+    await initCsrfToken();
+  }
+
   if (['POST', 'PUT', 'PATCH', 'DELETE'].indexOf(method) !== -1 && csrfToken) {
     options.headers['X-CSRF-Token'] = csrfToken;
   }
diff --git a/ccw/src/tools/claude-cli-tools.ts b/ccw/src/tools/claude-cli-tools.ts
index 9f5fedc1..922b6a2a 100644
--- a/ccw/src/tools/claude-cli-tools.ts
+++ b/ccw/src/tools/claude-cli-tools.ts
@@ -1,8 +1,9 @@
 /**
  * Claude CLI Tools Configuration Manager
- * Manages .claude/cli-tools.json with fallback:
- * 1. Project workspace: {projectDir}/.claude/cli-tools.json (priority)
- * 2. Global: ~/.claude/cli-tools.json (fallback)
+ * Manages .claude/cli-tools.json (tools) and .claude/cli-settings.json (settings)
+ * with fallback:
+ * 1. Project workspace: {projectDir}/.claude/ (priority)
+ * 2. Global: ~/.claude/ (fallback)
  */
 import * as fs from 'fs';
 import * as path from 'path';
@@ -15,6 +16,15 @@ export interface ClaudeCliTool {
   isBuiltin: boolean;
   command: string;
   description: string;
+  primaryModel?: string;
+  tags: string[];
+}
+
+export interface ClaudeCustomEndpoint {
+  id: string;
+  name: string;
+  enabled: boolean;
+  tags: string[];
 }
 
 export interface ClaudeCacheSettings {
@@ -23,76 +33,107 @@ export interface ClaudeCacheSettings {
   defaultSuffix: string;
 }
 
+// New: Tools-only config (cli-tools.json)
 export interface ClaudeCliToolsConfig {
   $schema?: string;
   version: string;
   tools: Record<string, ClaudeCliTool>;
-  customEndpoints: Array<{
-    id: string;
-    name: string;
-    enabled: boolean;
-  }>;
+  customEndpoints: ClaudeCustomEndpoint[];
+}
+
+// New: Settings-only config (cli-settings.json)
+export interface ClaudeCliSettingsConfig {
+  $schema?: string;
+  version: string;
   defaultTool: string;
-  settings: {
-    promptFormat: 'plain' | 'yaml' | 'json';
-    smartContext: {
-      enabled: boolean;
-      maxFiles: number;
+  promptFormat: 'plain' | 'yaml' | 'json';
+  smartContext: {
+    enabled: boolean;
+    maxFiles: number;
+  };
+  nativeResume: boolean;
+  recursiveQuery: boolean;
+  cache: ClaudeCacheSettings;
+  codeIndexMcp: 'codexlens' | 'ace' | 'none';
+}
+
+// Legacy combined config (for backward compatibility)
+export interface ClaudeCliCombinedConfig extends ClaudeCliToolsConfig {
+  defaultTool?: string;
+  settings?: {
+    promptFormat?: 'plain' | 'yaml' | 'json';
+    smartContext?: {
+      enabled?: boolean;
+      maxFiles?: number;
     };
-    nativeResume: boolean;
-    recursiveQuery: boolean;
-    cache: ClaudeCacheSettings;
-    codeIndexMcp: 'codexlens' | 'ace' | 'none';  // Code Index MCP provider
+    nativeResume?: boolean;
+    recursiveQuery?: boolean;
+    cache?: Partial<ClaudeCacheSettings>;
+    codeIndexMcp?: 'codexlens' | 'ace' | 'none';
   };
 }
 
 // ========== Default Config ==========
 
-const DEFAULT_CONFIG: ClaudeCliToolsConfig = {
-  version: '1.0.0',
+const DEFAULT_TOOLS_CONFIG: ClaudeCliToolsConfig = {
+  version: '2.0.0',
   tools: {
     gemini: {
       enabled: true,
       isBuiltin: true,
       command: 'gemini',
-      description: 'Google AI for code analysis'
+      description: 'Google AI for code analysis',
+      tags: []
     },
     qwen: {
       enabled: true,
       isBuiltin: true,
       command: 'qwen',
-      description: 'Alibaba AI assistant'
+      description: 'Alibaba AI assistant',
+      tags: []
     },
     codex: {
       enabled: true,
       isBuiltin: true,
       command: 'codex',
-      description: 'OpenAI code generation'
+      description: 'OpenAI code generation',
+      tags: []
     },
     claude: {
       enabled: true,
       isBuiltin: true,
       command: 'claude',
-      description: 'Anthropic AI assistant'
+      description: 'Anthropic AI assistant',
+      tags: []
+    },
+    opencode: {
+      enabled: true,
+      isBuiltin: true,
+      command: 'opencode',
+      description: 'OpenCode AI assistant',
+      primaryModel: 'opencode/glm-4.7-free',
+      tags: []
     }
   },
-  customEndpoints: [],
+  customEndpoints: []
+};
+
+const DEFAULT_SETTINGS_CONFIG: ClaudeCliSettingsConfig = {
+  version: '1.0.0',
   defaultTool: 'gemini',
-  settings: {
-    promptFormat: 'plain',
-    smartContext: {
-      enabled: false,
-      maxFiles: 10
-    },
-    nativeResume: true,
-    recursiveQuery: true,
-    cache: {
-      injectionMode: 'auto',
-      defaultPrefix: '',
-      defaultSuffix: ''
-    },
-    codeIndexMcp: 'codexlens'  // Default to CodexLens
-  }
+  promptFormat: 'plain',
+  smartContext: {
+    enabled: false,
+    maxFiles: 10
+  },
+  nativeResume: true,
+  recursiveQuery: true,
+  cache: {
+    injectionMode: 'auto',
+    defaultPrefix: '',
+    defaultSuffix: ''
+  },
+  codeIndexMcp: 'ace'
 };
 
 // ========== Helper Functions ==========
@@ -101,10 +142,18 @@ function getProjectConfigPath(projectDir: string): string {
   return path.join(projectDir, '.claude', 'cli-tools.json');
 }
 
+function getProjectSettingsPath(projectDir: string): string {
+  return path.join(projectDir, '.claude', 'cli-settings.json');
+}
+
 function getGlobalConfigPath(): string {
   return path.join(os.homedir(), '.claude', 'cli-tools.json');
 }
 
+function getGlobalSettingsPath(): string {
+  return path.join(os.homedir(), '.claude', 'cli-settings.json');
+}
+
 /**
  * Resolve config path with fallback:
  * 1. Project: {projectDir}/.claude/cli-tools.json
@@ -125,6 +174,25 @@ function resolveConfigPath(projectDir: string): { path: string; source: 'project
   return { path: projectPath, source: 'default' };
 }
 
+/**
+ * Resolve settings path with fallback:
+ * 1. Project: {projectDir}/.claude/cli-settings.json
+ * 2. Global: ~/.claude/cli-settings.json
+ */
+function resolveSettingsPath(projectDir: string): { path: string; source: 'project' | 'global' | 'default' } {
+  const projectPath = getProjectSettingsPath(projectDir);
+  if (fs.existsSync(projectPath)) {
+    return { path: projectPath, source: 'project' };
+  }
+
+  const globalPath = getGlobalSettingsPath();
+  if (fs.existsSync(globalPath)) {
+    return { path: globalPath, source: 'global' };
+  }
+
+  return { path: projectPath, source: 'default' };
+}
+
 function ensureClaudeDir(projectDir: string): void {
   const claudeDir = path.join(projectDir, '.claude');
   if (!fs.existsSync(claudeDir)) {
@@ -134,6 +202,20 @@ function ensureClaudeDir(projectDir: string): void {
 
 // ========== Main Functions ==========
 
+/**
+ * Ensure tool has tags field (for backward compatibility)
+ */
+function ensureToolTags(tool: Partial<ClaudeCliTool>): ClaudeCliTool {
+  return {
+    enabled: tool.enabled ?? true,
+    isBuiltin: tool.isBuiltin ?? false,
+    command: tool.command ?? '',
+    description: tool.description ?? '',
+    primaryModel: tool.primaryModel,
+    tags: tool.tags ?? []
+  };
+}
+
 /**
  * Load CLI tools configuration with fallback:
  * 1. Project: {projectDir}/.claude/cli-tools.json
@@ -145,61 +227,115 @@ export function loadClaudeCliTools(projectDir: string): ClaudeCliToolsConfig & {
 
   try {
     if (resolved.source === 'default') {
-      // No config file found, return defaults
-      return { ...DEFAULT_CONFIG, _source: 'default' };
+      return { ...DEFAULT_TOOLS_CONFIG, _source: 'default' };
     }
 
     const content = fs.readFileSync(resolved.path, 'utf-8');
-    const parsed = JSON.parse(content) as Partial<ClaudeCliToolsConfig>;
+    const parsed = JSON.parse(content) as Partial<ClaudeCliCombinedConfig>;
 
-    // Merge with defaults
-    const config = {
-      ...DEFAULT_CONFIG,
-      ...parsed,
-      tools: { ...DEFAULT_CONFIG.tools, ...(parsed.tools || {}) },
-      settings: {
-        ...DEFAULT_CONFIG.settings,
-        ...(parsed.settings || {}),
-        smartContext: {
-          ...DEFAULT_CONFIG.settings.smartContext,
-          ...(parsed.settings?.smartContext || {})
-        },
-        cache: {
-          ...DEFAULT_CONFIG.settings.cache,
-          ...(parsed.settings?.cache || {})
-        }
-      },
+    // Merge tools with defaults and ensure tags exist
+    const mergedTools: Record<string, ClaudeCliTool> = {};
+    for (const [key, tool] of Object.entries({ ...DEFAULT_TOOLS_CONFIG.tools, ...(parsed.tools || {}) })) {
+      mergedTools[key] = ensureToolTags(tool);
+    }
+
+    // Ensure customEndpoints have tags
+    const mergedEndpoints = (parsed.customEndpoints || []).map(ep => ({
+      ...ep,
+      tags: ep.tags ?? []
+    }));
+
+    const config: ClaudeCliToolsConfig & { _source?: string } = {
+      version: parsed.version || DEFAULT_TOOLS_CONFIG.version,
+      tools: mergedTools,
+      customEndpoints: mergedEndpoints,
+      $schema: parsed.$schema,
       _source: resolved.source
     };
 
-    console.log(`[claude-cli-tools] Loaded config from ${resolved.source}: ${resolved.path}`);
+    console.log(`[claude-cli-tools] Loaded tools config from ${resolved.source}: ${resolved.path}`);
     return config;
   } catch (err) {
-    console.error('[claude-cli-tools] Error loading config:', err);
-    return { ...DEFAULT_CONFIG, _source: 'default' };
+    console.error('[claude-cli-tools] Error loading tools config:', err);
+    return { ...DEFAULT_TOOLS_CONFIG, _source: 'default' };
   }
 }
 
 /**
  * Save CLI tools configuration to project .claude/cli-tools.json
- * Always saves to project directory (not global)
  */
 export function saveClaudeCliTools(projectDir: string, config: ClaudeCliToolsConfig & { _source?: string }): void {
   ensureClaudeDir(projectDir);
   const configPath = getProjectConfigPath(projectDir);
 
-  // Remove internal _source field before saving
   const { _source, ...configToSave } = config;
 
   try {
     fs.writeFileSync(configPath, JSON.stringify(configToSave, null, 2), 'utf-8');
-    console.log(`[claude-cli-tools] Saved config to project: ${configPath}`);
+    console.log(`[claude-cli-tools] Saved tools config to: ${configPath}`);
   } catch (err) {
-    console.error('[claude-cli-tools] Error saving config:', err);
+    console.error('[claude-cli-tools] Error saving tools config:', err);
     throw new Error(`Failed to save CLI tools config: ${err}`);
   }
 }
 
+/**
+ * Load CLI settings configuration with fallback:
+ * 1. Project: {projectDir}/.claude/cli-settings.json
+ * 2. Global: ~/.claude/cli-settings.json
+ * 3. Default settings
+ */
+export function loadClaudeCliSettings(projectDir: string): ClaudeCliSettingsConfig & { _source?: string } {
+  const resolved = resolveSettingsPath(projectDir);
+
+  try {
+    if (resolved.source === 'default') {
+      return { ...DEFAULT_SETTINGS_CONFIG, _source: 'default' };
+    }
+
+    const content = fs.readFileSync(resolved.path, 'utf-8');
+    const parsed = JSON.parse(content) as Partial<ClaudeCliSettingsConfig>;
+
+    const config: ClaudeCliSettingsConfig & { _source?: string } = {
+      ...DEFAULT_SETTINGS_CONFIG,
+      ...parsed,
+      smartContext: {
+        ...DEFAULT_SETTINGS_CONFIG.smartContext,
+        ...(parsed.smartContext || {})
+      },
+      cache: {
+        ...DEFAULT_SETTINGS_CONFIG.cache,
+        ...(parsed.cache || {})
+      },
+      _source: resolved.source
+    };
+
+    console.log(`[claude-cli-tools] Loaded settings from ${resolved.source}: ${resolved.path}`);
+    return config;
+  } catch (err) {
+    console.error('[claude-cli-tools] Error loading settings:', err);
+    return { ...DEFAULT_SETTINGS_CONFIG, _source: 'default' };
+  }
+}
+
+/**
+ * Save CLI settings configuration to project .claude/cli-settings.json
+ */
+export function saveClaudeCliSettings(projectDir: string, config: ClaudeCliSettingsConfig & { _source?: string }): void {
+  ensureClaudeDir(projectDir);
+  const settingsPath = getProjectSettingsPath(projectDir);
+
+  const { _source, ...configToSave } = config;
+
+  try {
+    fs.writeFileSync(settingsPath, JSON.stringify(configToSave, null, 2), 'utf-8');
+    console.log(`[claude-cli-tools] Saved settings to: ${settingsPath}`);
+  } catch (err) {
+    console.error('[claude-cli-tools] Error saving settings:', err);
+    throw new Error(`Failed to save CLI settings: ${err}`);
+  }
+}
+
 /**
  * Update enabled status for a specific tool
  */
@@ -224,16 +360,16 @@ export function updateClaudeToolEnabled(
 export function updateClaudeCacheSettings(
   projectDir: string,
   cacheSettings: Partial<ClaudeCacheSettings>
-): ClaudeCliToolsConfig {
-  const config = loadClaudeCliTools(projectDir);
+): ClaudeCliSettingsConfig {
+  const settings = loadClaudeCliSettings(projectDir);
 
-  config.settings.cache = {
-    ...config.settings.cache,
+  settings.cache = {
+    ...settings.cache,
     ...cacheSettings
   };
 
-  saveClaudeCliTools(projectDir, config);
-  return config;
+  saveClaudeCliSettings(projectDir, settings);
+  return settings;
 }
 
 /**
@@ -242,11 +378,11 @@ export function updateClaudeCacheSettings(
 export function updateClaudeDefaultTool(
   projectDir: string,
   defaultTool: string
-): ClaudeCliToolsConfig {
-  const config = loadClaudeCliTools(projectDir);
-  config.defaultTool = defaultTool;
-  saveClaudeCliTools(projectDir, config);
-  return config;
+): ClaudeCliSettingsConfig {
+  const settings = loadClaudeCliSettings(projectDir);
+  settings.defaultTool = defaultTool;
+  saveClaudeCliSettings(projectDir, settings);
+  return settings;
 }
 
 /**
@@ -254,16 +390,23 @@ export function updateClaudeDefaultTool(
  */
 export function addClaudeCustomEndpoint(
   projectDir: string,
-  endpoint: { id: string; name: string; enabled: boolean }
+  endpoint: { id: string; name: string; enabled: boolean; tags?: string[] }
 ): ClaudeCliToolsConfig {
   const config = loadClaudeCliTools(projectDir);
 
+  const newEndpoint: ClaudeCustomEndpoint = {
+    id: endpoint.id,
+    name: endpoint.name,
+    enabled: endpoint.enabled,
+    tags: endpoint.tags || []
+  };
+
   // Check if endpoint already exists
   const existingIndex = config.customEndpoints.findIndex(e => e.id === endpoint.id);
   if (existingIndex >= 0) {
-    config.customEndpoints[existingIndex] = endpoint;
+    config.customEndpoints[existingIndex] = newEndpoint;
   } else {
-    config.customEndpoints.push(endpoint);
+    config.customEndpoints.push(newEndpoint);
   }
 
   saveClaudeCliTools(projectDir, config);
@@ -309,12 +452,12 @@ export function getClaudeCliToolsInfo(projectDir: string): {
 export function updateCodeIndexMcp(
   projectDir: string,
   provider: 'codexlens' | 'ace' | 'none'
-): { success: boolean; error?: string; config?: ClaudeCliToolsConfig } {
+): { success: boolean; error?: string; settings?: ClaudeCliSettingsConfig } {
   try {
-    // Update config
-    const config = loadClaudeCliTools(projectDir);
-    config.settings.codeIndexMcp = provider;
-    saveClaudeCliTools(projectDir, config);
+    // Update settings config
+    const settings = loadClaudeCliSettings(projectDir);
+    settings.codeIndexMcp = provider;
+    saveClaudeCliSettings(projectDir, settings);
 
     // Only update global CLAUDE.md (consistent with Chinese response / Windows platform)
     const globalClaudeMdPath = path.join(os.homedir(), '.claude', 'CLAUDE.md');
@@ -358,7 +501,7 @@ export function updateCodeIndexMcp(
       console.log(`[claude-cli-tools] Updated global CLAUDE.md to use ${provider}`);
     }
 
-    return { success: true, config };
+    return { success: true, settings };
   } catch (err) {
     console.error('[claude-cli-tools] Error updating Code Index MCP:', err);
     return { success: false, error: (err as Error).message };
@@ -369,8 +512,8 @@ export function updateCodeIndexMcp(
  * Get current Code Index MCP provider
  */
 export function getCodeIndexMcp(projectDir: string): 'codexlens' | 'ace' | 'none' {
-  const config = loadClaudeCliTools(projectDir);
-  return config.settings.codeIndexMcp || 'codexlens';
+  const settings = loadClaudeCliSettings(projectDir);
+  return settings.codeIndexMcp || 'ace';
 }
 
 /**
diff --git a/ccw/src/tools/cli-config-manager.ts b/ccw/src/tools/cli-config-manager.ts
index 876a72ea..b94e3e74 100644
--- a/ccw/src/tools/cli-config-manager.ts
+++ b/ccw/src/tools/cli-config-manager.ts
@@ -6,6 +6,7 @@
 import * as fs from 'fs';
 import * as path from 'path';
 import { StoragePaths, ensureStorageDir } from '../config/storage-paths.js';
+import { loadClaudeCliTools, saveClaudeCliTools } from './claude-cli-tools.js';
 
 // ========== Types ==========
 
@@ -234,6 +235,20 @@ export function updateToolConfig(
   config.tools[tool] = updatedToolConfig;
   saveCliConfig(baseDir, config);
 
+  // Also sync tags to cli-tools.json
+  if (updates.tags !== undefined) {
+    try {
+      const claudeCliTools = loadClaudeCliTools(baseDir);
+      if (claudeCliTools.tools[tool]) {
+        claudeCliTools.tools[tool].tags = updatedToolConfig.tags || [];
+        saveClaudeCliTools(baseDir, claudeCliTools);
+      }
+    } catch (err) {
+      // Log warning instead of ignoring errors syncing to cli-tools.json
+      console.warn(`[cli-config] Failed to sync tags to cli-tools.json for tool '${tool}'.`, err);
+    }
+  }
+
   return updatedToolConfig;
 }
 
@@ -298,14 +313,30 @@ export function getPredefinedModels(tool: string): string[] {
 }
 
 /**
- * Get full config response for API (includes predefined models)
+ * Get full config response for API (includes predefined models and tags from cli-tools.json)
  */
 export function getFullConfigResponse(baseDir: string): {
   config: CliConfig;
   predefinedModels: Record<string, string[]>;
 } {
+  const config = loadCliConfig(baseDir);
+
+  // Merge tags from cli-tools.json
+  try {
+    const claudeCliTools = loadClaudeCliTools(baseDir);
+    for (const [toolName, toolConfig] of Object.entries(config.tools)) {
+      const claudeTool = claudeCliTools.tools[toolName];
+      if (claudeTool && claudeTool.tags) {
+        toolConfig.tags = claudeTool.tags;
+      }
+    }
+  } catch (err) {
+    // Log warning instead of ignoring errors loading cli-tools.json
+    console.warn('[cli-config] Could not merge tags from cli-tools.json.', err);
+  }
+
   return {
-    config: loadCliConfig(baseDir),
+    config,
     predefinedModels: { ...PREDEFINED_MODELS }
   };
 }
diff --git a/ccw/src/types/litellm-api-config.ts b/ccw/src/types/litellm-api-config.ts
index 4fcf1ebb..5d77baa7 100644
--- a/ccw/src/types/litellm-api-config.ts
+++ b/ccw/src/types/litellm-api-config.ts
@@ -353,6 +353,7 @@ export interface CodexLensEmbeddingRotation {
 /**
  * Generic embedding pool configuration (refactored from CodexLensEmbeddingRotation)
  * Supports automatic discovery of all providers offering a specific model
+ * @deprecated Use ModelPoolConfig instead
  */
 export interface EmbeddingPoolConfig {
   /** Whether embedding pool is enabled */
@@ -377,6 +378,50 @@ export interface EmbeddingPoolConfig {
   defaultMaxConcurrentPerKey: number;
 }
 
+/**
+ * Model type for pool configuration
+ */
+export type ModelPoolType = 'embedding' | 'llm' | 'reranker';
+
+/**
+ * Individual model pool configuration
+ * Supports embedding, LLM, and reranker models with high availability
+ */
+export interface ModelPoolConfig {
+  /** Unique identifier for this pool */
+  id: string;
+
+  /** Model type: embedding, llm, or reranker */
+  modelType: ModelPoolType;
+
+  /** Whether this pool is enabled */
+  enabled: boolean;
+
+  /** Target model name (e.g., "text-embedding-3-small", "gpt-4o") */
+  targetModel: string;
+
+  /** Selection strategy: round_robin, latency_aware, weighted_random */
+  strategy: 'round_robin' | 'latency_aware' | 'weighted_random';
+
+  /** Whether to automatically discover all providers offering targetModel */
+  autoDiscover: boolean;
+
+  /** Provider IDs to exclude from auto-discovery (optional) */
+  excludedProviderIds?: string[];
+
+  /** Default cooldown seconds for rate-limited endpoints (default: 60) */
+  defaultCooldown: number;
+
+  /** Default maximum concurrent requests per key (default: 4) */
+  defaultMaxConcurrentPerKey: number;
+
+  /** Optional display name for this pool */
+  name?: string;
+
+  /** Optional description */
+  description?: string;
+}
+
 /**
  * Complete LiteLLM API configuration
  * Root configuration object stored in JSON file
@@ -400,6 +445,9 @@ export interface LiteLLMApiConfig {
   /** CodexLens multi-provider embedding rotation config (deprecated, use embeddingPoolConfig) */
   codexlensEmbeddingRotation?: CodexLensEmbeddingRotation;
 
-  /** Generic embedding pool configuration with auto-discovery support */
+  /** Generic embedding pool configuration with auto-discovery support (deprecated, use modelPools) */
   embeddingPoolConfig?: EmbeddingPoolConfig;
+
+  /** Multi-model pool configurations (supports embedding, LLM, reranker) */
+  modelPools?: ModelPoolConfig[];
 }