diff --git a/server/src/llm/capabilities.ts b/server/src/llm/capabilities.ts
index fc6e718f2..c7e292d54 100644
--- a/server/src/llm/capabilities.ts
+++ b/server/src/llm/capabilities.ts
@@ -160,6 +160,10 @@ export function getJsonCapability(provider: LLMProvider, model?: string, baseURL
       supportsJsonObject: false,
       supportsJsonSchema: false,
     },
+    atlascloud: {
+      supportsJsonObject: false,
+      supportsJsonSchema: false,
+    },
   };
 
   const cap = isBuiltinLLMProvider(provider) ? jsonCapabilities[provider] : undefined;
diff --git a/server/src/llm/providers.ts b/server/src/llm/providers.ts
index 6e6ee85fb..6d628eb4d 100644
--- a/server/src/llm/providers.ts
+++ b/server/src/llm/providers.ts
@@ -11,8 +11,11 @@ export interface ProviderConfig {
   defaultModel: string;
   models: string[];
   envKey: string;
+  envKeyAliases?: string[];
   envBaseURLKey?: string;
+  envBaseURLKeyAliases?: string[];
   envModelKey?: string;
+  envModelKeyAliases?: string[];
   maxTokens?: number;
   requiresApiKey?: boolean;
 }
@@ -140,6 +143,18 @@ export const PROVIDERS: Record<BuiltinLLMProvider, ProviderConfig> = {
     envModelKey: "OLLAMA_MODEL",
     requiresApiKey: false,
   },
+  atlascloud: {
+    name: "Atlas Cloud",
+    baseURL: "https://api.atlascloud.ai/v1",
+    defaultModel: "qwen/qwen3.5-flash",
+    models: ["qwen/qwen3.5-flash", "deepseek-ai/deepseek-v4-pro"],
+    envKey: "ATLASCLOUD_API_KEY",
+    envKeyAliases: ["ATLAS_CLOUD_API_KEY"],
+    envBaseURLKey: "ATLASCLOUD_BASE_URL",
+    envBaseURLKeyAliases: ["ATLAS_CLOUD_BASE_URL"],
+    envModelKey: "ATLASCLOUD_MODEL",
+    envModelKeyAliases: ["ATLAS_CLOUD_MODEL"],
+  },
 };
 
 export const SUPPORTED_PROVIDERS: BuiltinLLMProvider[] = [...LLM_PROVIDERS];
@@ -152,37 +167,42 @@ export function normalizeBaseURL(baseURL: string): string {
   return baseURL.endsWith("/") ? baseURL.slice(0, -1) : baseURL;
 }
 
+function readProviderEnv(keys: Array<string | undefined>, normalize?: (value: string) => string): string | undefined {
+  for (const key of keys) {
+    if (!key) {
+      continue;
+    }
+    const value = process.env[key];
+    if (typeof value === "string" && value.trim()) {
+      const trimmed = value.trim();
+      return normalize ? normalize(trimmed) : trimmed;
+    }
+  }
+  return undefined;
+}
+
 export function getProviderEnvApiKey(provider: LLMProvider): string | undefined {
   if (!isBuiltInProvider(provider)) {
     return undefined;
   }
-  const envKey = PROVIDERS[provider].envKey;
-  const value = process.env[envKey];
-  return typeof value === "string" && value.trim() ? value.trim() : undefined;
+  const config = PROVIDERS[provider];
+  return readProviderEnv([config.envKey, ...(config.envKeyAliases ?? [])]);
 }
 
 export function getProviderEnvBaseUrl(provider: LLMProvider): string | undefined {
   if (!isBuiltInProvider(provider)) {
     return undefined;
   }
-  const envKey = PROVIDERS[provider].envBaseURLKey;
-  if (!envKey) {
-    return undefined;
-  }
-  const value = process.env[envKey];
-  return typeof value === "string" && value.trim() ? normalizeBaseURL(value.trim()) : undefined;
+  const config = PROVIDERS[provider];
+  return readProviderEnv([config.envBaseURLKey, ...(config.envBaseURLKeyAliases ?? [])], normalizeBaseURL);
 }
 
 export function getProviderEnvModel(provider: LLMProvider): string | undefined {
   if (!isBuiltInProvider(provider)) {
     return undefined;
   }
-  const envKey = PROVIDERS[provider].envModelKey;
-  if (!envKey) {
-    return undefined;
-  }
-  const value = process.env[envKey];
-  return typeof value === "string" && value.trim() ? value.trim() : undefined;
+  const config = PROVIDERS[provider];
+  return readProviderEnv([config.envModelKey, ...(config.envModelKeyAliases ?? [])]);
 }
 
 export function getProviderDefaultBaseUrl(provider: LLMProvider): string | undefined {
diff --git a/server/src/services/settings/RagSettingsService.ts b/server/src/services/settings/RagSettingsService.ts
index 3672cfc1b..b2bceee01 100644
--- a/server/src/services/settings/RagSettingsService.ts
+++ b/server/src/services/settings/RagSettingsService.ts
@@ -1,3 +1,4 @@
+import type { BuiltinLLMProvider } from "@ai-novel/shared/types/llm";
 import { prisma } from "../../db/prisma";
 import { ragConfig, asEmbeddingProvider, type EmbeddingProvider } from "../../config/rag";
 import {
@@ -76,6 +77,12 @@ export interface SaveRagEmbeddingSettingsResult {
   shouldReindex: boolean;
 }
 
+const NON_EMBEDDING_BUILT_IN_PROVIDERS = new Set<BuiltinLLMProvider>(["atlascloud"]);
+
+export function getBuiltInRagEmbeddingProviders(): BuiltinLLMProvider[] {
+  return SUPPORTED_PROVIDERS.filter((provider) => !NON_EMBEDDING_BUILT_IN_PROVIDERS.has(provider));
+}
+
 function normalizeEmbeddingModel(value: string | undefined, provider: EmbeddingProvider = ragConfig.embeddingProvider): string {
   const normalized = value?.trim();
   if (normalized && normalized.length > 0) {
@@ -394,7 +401,7 @@ export async function saveRagEmbeddingSettings(input: RagEmbeddingSettingsInput)
 }
 
 export async function getRagEmbeddingProviders(): Promise<RagEmbeddingProviderStatus[]> {
-  const builtInProviders = [...SUPPORTED_PROVIDERS];
+  const builtInProviders = getBuiltInRagEmbeddingProviders();
   try {
     const items = await prisma.aPIKey.findMany({
       select: {
diff --git a/server/tests/llmProviders.test.js b/server/tests/llmProviders.test.js
index 14c21e9f2..95f74bd6e 100644
--- a/server/tests/llmProviders.test.js
+++ b/server/tests/llmProviders.test.js
@@ -1,7 +1,14 @@
 const test = require("node:test");
 const assert = require("node:assert/strict");
 const { z } = require("zod");
-const { PROVIDERS, SUPPORTED_PROVIDERS } = require("../dist/llm/providers.js");
+const {
+  getProviderEnvApiKey,
+  getProviderEnvBaseUrl,
+  getProviderEnvModel,
+  PROVIDERS,
+  resolveProviderBaseUrl,
+  SUPPORTED_PROVIDERS,
+} = require("../dist/llm/providers.js");
 const {
   getJsonCapability,
   getModelParameterCompatibility,
@@ -14,14 +21,14 @@ const {
   selectStructuredOutputStrategy,
 } = require("../dist/llm/structuredOutput.js");
 
-test("supported providers include kimi, minimax, glm, qwen, gemini and ollama", () => {
-  for (const provider of ["kimi", "minimax", "glm", "qwen", "gemini", "ollama"]) {
+test("supported providers include kimi, minimax, glm, qwen, gemini, ollama and atlascloud", () => {
+  for (const provider of ["kimi", "minimax", "glm", "qwen", "gemini", "ollama", "atlascloud"]) {
     assert.ok(SUPPORTED_PROVIDERS.includes(provider), `${provider} should be available`);
   }
 });
 
 test("new provider defaults are present in their model fallback lists", () => {
-  for (const provider of ["kimi", "minimax", "glm", "qwen", "gemini", "ollama"]) {
+  for (const provider of ["kimi", "minimax", "glm", "qwen", "gemini", "ollama", "atlascloud"]) {
     assert.ok(
       PROVIDERS[provider].models.includes(PROVIDERS[provider].defaultModel),
       `${provider} default model should exist in fallback models`,
@@ -29,6 +36,50 @@ test("new provider defaults are present in their model fallback lists", () => {
   }
 });
 
+test("atlascloud defaults and environment aliases use the OpenAI-compatible endpoint", () => {
+  const envKeys = [
+    "ATLASCLOUD_API_KEY",
+    "ATLAS_CLOUD_API_KEY",
+    "ATLASCLOUD_BASE_URL",
+    "ATLAS_CLOUD_BASE_URL",
+    "ATLASCLOUD_MODEL",
+    "ATLAS_CLOUD_MODEL",
+  ];
+  const previous = Object.fromEntries(envKeys.map((key) => [key, process.env[key]]));
+
+  try {
+    for (const key of envKeys) {
+      delete process.env[key];
+    }
+
+    assert.equal(PROVIDERS.atlascloud.name, "Atlas Cloud");
+    assert.equal(PROVIDERS.atlascloud.baseURL, "https://api.atlascloud.ai/v1");
+    assert.equal(PROVIDERS.atlascloud.defaultModel, "qwen/qwen3.5-flash");
+    assert.ok(PROVIDERS.atlascloud.models.includes("deepseek-ai/deepseek-v4-pro"));
+    assert.equal(resolveProviderBaseUrl("atlascloud"), "https://api.atlascloud.ai/v1");
+
+    process.env.ATLAS_CLOUD_API_KEY = " alias-key ";
+    assert.equal(getProviderEnvApiKey("atlascloud"), "alias-key");
+    process.env.ATLASCLOUD_API_KEY = " primary-key ";
+    assert.equal(getProviderEnvApiKey("atlascloud"), "primary-key");
+
+    process.env.ATLAS_CLOUD_BASE_URL = "https://proxy.example/v1/";
+    assert.equal(getProviderEnvBaseUrl("atlascloud"), "https://proxy.example/v1");
+    assert.equal(resolveProviderBaseUrl("atlascloud"), "https://proxy.example/v1");
+
+    process.env.ATLAS_CLOUD_MODEL = " deepseek-ai/deepseek-v4-pro ";
+    assert.equal(getProviderEnvModel("atlascloud"), "deepseek-ai/deepseek-v4-pro");
+  } finally {
+    for (const key of envKeys) {
+      if (previous[key] === undefined) {
+        delete process.env[key];
+      } else {
+        process.env[key] = previous[key];
+      }
+    }
+  }
+});
+
 test("kimi thinking models do not enable forced json mode", () => {
   const stableCapability = getJsonCapability("kimi", "moonshot-v1-32k");
   assert.equal(stableCapability.supportsJsonObject, true);
@@ -118,6 +169,16 @@ test("structured output profiles distinguish official, ModelScope Qwen and unkno
   assert.equal(qwenBehindProxyProfile.nativeJsonSchema, false);
   assert.equal(selectStructuredOutputStrategy(qwenBehindProxyProfile, schema), "prompt_json");
 
+  const atlasCloudProfile = resolveStructuredOutputProfile({
+    provider: "atlascloud",
+    model: "qwen/qwen3.5-flash",
+    baseURL: "https://api.atlascloud.ai/v1",
+    executionMode: "structured",
+  });
+  assert.equal(atlasCloudProfile.family, "custom_openai_compatible_qwen");
+  assert.equal(atlasCloudProfile.nativeJsonSchema, false);
+  assert.equal(selectStructuredOutputStrategy(atlasCloudProfile, schema), "prompt_json");
+
   const deepseekBehindProxyProfile = resolveStructuredOutputProfile({
     provider: "openai",
     model: "deepseek-chat",
@@ -260,6 +321,19 @@ test("resolveLLMClientOptions applies structured reasoning and token guardrails"
     assert.equal(qwenThinking.maxTokens, 8192);
     assert.equal(qwenThinking.requestProtocol, "openai_compatible");
 
+    const atlasCloud = await resolveLLMClientOptions("atlascloud", {
+      apiKey: "test-key",
+      executionMode: "structured",
+      structuredStrategy: "prompt_json",
+      maxTokens: 20000,
+    });
+    assert.equal(atlasCloud.providerName, "Atlas Cloud");
+    assert.equal(atlasCloud.model, "qwen/qwen3.5-flash");
+    assert.equal(atlasCloud.baseURL, "https://api.atlascloud.ai/v1");
+    assert.equal(atlasCloud.structuredProfile?.family, "custom_openai_compatible_qwen");
+    assert.equal(atlasCloud.maxTokens, 8192);
+    assert.equal(atlasCloud.requestProtocol, "openai_compatible");
+
     const anthropicProtocol = await resolveLLMClientOptions("openai", {
       apiKey: "test-key",
       model: "claude-sonnet-4-5",
diff --git a/shared/types/llm.ts b/shared/types/llm.ts
index a15eeb6be..5c240dac7 100644
--- a/shared/types/llm.ts
+++ b/shared/types/llm.ts
@@ -10,6 +10,7 @@ export const LLM_PROVIDERS = [
   "qwen",
   "gemini",
   "ollama",
+  "atlascloud",
 ] as const;
 
 export type BuiltinLLMProvider = typeof LLM_PROVIDERS[number];