diff --git a/docs/01-getting-started.md b/docs/01-getting-started.md
index 0ea3624df..713a04ad7 100644
--- a/docs/01-getting-started.md
+++ b/docs/01-getting-started.md
@@ -253,6 +253,8 @@ OpenRouter is a universal AI API gateway — one account gives you access to eve
 
 Why OpenRouter instead of OpenAI directly? One account, one key, one billing relationship — and it future-proofs you for Claude, Gemini, or any other model later.
 
+Atlas Cloud is also supported as an OpenAI-compatible backend for the core MCP server. To use it instead of OpenRouter, set `OPEN_BRAIN_AI_PROVIDER=atlascloud` and `ATLASCLOUD_API_KEY`, and keep the embedding dimension at 1536 unless you also update the database vector column.
+
 1. Go to [openrouter.ai](https://openrouter.ai) and sign up
 2. Go to [openrouter.ai/keys](https://openrouter.ai/keys)
 3. Click **Create Key**, name it `open-brain`
@@ -412,6 +414,12 @@ Set your OpenRouter key from Step 4:
 supabase secrets set OPENROUTER_API_KEY=your-openrouter-key-here
 ```
 
+Or, to use Atlas Cloud instead:
+
+```bash
+supabase secrets set OPEN_BRAIN_AI_PROVIDER=atlascloud ATLASCLOUD_API_KEY=your-atlascloud-key-here
+```
+
 > [!NOTE]
 > `SUPABASE_URL` and `SUPABASE_SERVICE_ROLE_KEY` are automatically available inside Edge Functions — you don't need to set them.
 >
@@ -594,6 +602,12 @@ Set your OpenRouter key from Step 4:
 supabase secrets set OPENROUTER_API_KEY=your-openrouter-key-here
 ```
 
+Or, to use Atlas Cloud instead:
+
+```powershell
+supabase secrets set OPEN_BRAIN_AI_PROVIDER=atlascloud ATLASCLOUD_API_KEY=your-atlascloud-key-here
+```
+
 > [!NOTE]
 > `SUPABASE_URL` and `SUPABASE_SERVICE_ROLE_KEY` are automatically available inside Edge Functions — you don't need to set them.
 >
@@ -912,7 +926,7 @@ The embedding is what makes retrieval powerful. "Sarah's thinking about leaving"
 
 ### Swapping Models Later
 
-Because you're using OpenRouter, you can swap models by editing the model strings in the Edge Function code and redeploying. Browse available models at [openrouter.ai/models](https://openrouter.ai/models). Just make sure embedding dimensions match (1536 for the current setup).
+Because the server uses OpenAI-compatible endpoints, you can swap models by setting `OPEN_BRAIN_CHAT_MODEL` or `OPEN_BRAIN_EMBEDDING_MODEL` before redeploying. Browse OpenRouter models at [openrouter.ai/models](https://openrouter.ai/models), or use Atlas Cloud model IDs such as `qwen/qwen3.5-flash` for chat. Just make sure embedding dimensions match (1536 for the current setup).
 
 </details>
 
diff --git a/docs/open-brain-assistant-gpt-context.md b/docs/open-brain-assistant-gpt-context.md
index f8b4491a3..4057d6dad 100644
--- a/docs/open-brain-assistant-gpt-context.md
+++ b/docs/open-brain-assistant-gpt-context.md
@@ -88,15 +88,16 @@ For Claude Desktop and ChatGPT, the `?key=` URL is usually the simplest path. Fo
 Open Brain setup requires:
 
 - Supabase project.
-- OpenRouter account and API key.
+- OpenRouter account and API key, or Atlas Cloud configured as the OpenAI-compatible backend.
 - Supabase CLI.
 - A generated MCP access key.
 - pgvector enabled in Supabase.
-- `OPENROUTER_API_KEY` and `MCP_ACCESS_KEY` set as Supabase Edge Function secrets.
+- `OPENROUTER_API_KEY` and `MCP_ACCESS_KEY` set as Supabase Edge Function secrets. For Atlas Cloud, use `OPEN_BRAIN_AI_PROVIDER=atlascloud` plus `ATLASCLOUD_API_KEY` instead of `OPENROUTER_API_KEY`.
 
 Important details:
 
 - `SUPABASE_URL` and `SUPABASE_SERVICE_ROLE_KEY` are automatically available inside Supabase Edge Functions.
+- OpenRouter remains the default AI backend. Atlas Cloud is selected explicitly with `OPEN_BRAIN_AI_PROVIDER=atlascloud`, or automatically when `ATLASCLOUD_API_KEY` is set and `OPENROUTER_API_KEY` is absent.
 - Newer Supabase projects may not grant service role table permissions by default. The user must run the documented `GRANT` SQL for the `thoughts` table.
 - The base embedding dimension is 1536. If users change embedding models, dimensions must still match the vector column and search function.
 - The `upsert_thought` function deduplicates by normalized content fingerprint and merges metadata on duplicate capture.
diff --git a/server/index.ts b/server/index.ts
index d3265f958..a9cb7e3e3 100644
--- a/server/index.ts
+++ b/server/index.ts
@@ -8,10 +8,75 @@ import { createClient } from "@supabase/supabase-js";
 
 const SUPABASE_URL = Deno.env.get("SUPABASE_URL")!;
 const SUPABASE_SERVICE_ROLE_KEY = Deno.env.get("SUPABASE_SERVICE_ROLE_KEY")!;
-const OPENROUTER_API_KEY = Deno.env.get("OPENROUTER_API_KEY")!;
 const MCP_ACCESS_KEY = Deno.env.get("MCP_ACCESS_KEY")!;
 
 const OPENROUTER_BASE = "https://openrouter.ai/api/v1";
+const ATLASCLOUD_BASE = "https://api.atlascloud.ai/v1";
+const DEFAULT_OPENROUTER_EMBEDDING_MODEL = "openai/text-embedding-3-small";
+const DEFAULT_OPENROUTER_CHAT_MODEL = "openai/gpt-4o-mini";
+const DEFAULT_ATLASCLOUD_EMBEDDING_MODEL = "text-embedding-3-small";
+const DEFAULT_ATLASCLOUD_CHAT_MODEL = "qwen/qwen3.5-flash";
+
+const OPENROUTER_API_KEY = Deno.env.get("OPENROUTER_API_KEY");
+const ATLASCLOUD_API_KEY =
+  Deno.env.get("ATLASCLOUD_API_KEY") || Deno.env.get("ATLAS_CLOUD_API_KEY");
+
+type AiProvider = {
+  name: "OpenRouter" | "Atlas Cloud";
+  baseUrl: string;
+  apiKey: string;
+  embeddingModel: string;
+  chatModel: string;
+};
+
+function trimTrailingSlash(url: string): string {
+  return url.replace(/\/+$/, "");
+}
+
+function buildAiProviderConfig(): AiProvider {
+  const requestedProvider = Deno.env.get("OPEN_BRAIN_AI_PROVIDER")?.toLowerCase();
+  const useAtlasCloud =
+    requestedProvider === "atlascloud" || requestedProvider === "atlas_cloud" ||
+    (!OPENROUTER_API_KEY && !!ATLASCLOUD_API_KEY);
+
+  if (useAtlasCloud) {
+    if (!ATLASCLOUD_API_KEY) {
+      throw new Error(
+        "Atlas Cloud selected but ATLASCLOUD_API_KEY or ATLAS_CLOUD_API_KEY is not set"
+      );
+    }
+
+    return {
+      name: "Atlas Cloud",
+      baseUrl: trimTrailingSlash(
+        Deno.env.get("ATLASCLOUD_API_BASE") ||
+          Deno.env.get("ATLAS_CLOUD_API_BASE") ||
+          ATLASCLOUD_BASE
+      ),
+      apiKey: ATLASCLOUD_API_KEY,
+      embeddingModel:
+        Deno.env.get("OPEN_BRAIN_EMBEDDING_MODEL") || DEFAULT_ATLASCLOUD_EMBEDDING_MODEL,
+      chatModel: Deno.env.get("OPEN_BRAIN_CHAT_MODEL") || DEFAULT_ATLASCLOUD_CHAT_MODEL,
+    };
+  }
+
+  if (!OPENROUTER_API_KEY) {
+    throw new Error(
+      "OPENROUTER_API_KEY is not set. Set it, or set OPEN_BRAIN_AI_PROVIDER=atlascloud with ATLASCLOUD_API_KEY."
+    );
+  }
+
+  return {
+    name: "OpenRouter",
+    baseUrl: trimTrailingSlash(Deno.env.get("OPENROUTER_BASE_URL") || OPENROUTER_BASE),
+    apiKey: OPENROUTER_API_KEY,
+    embeddingModel:
+      Deno.env.get("OPEN_BRAIN_EMBEDDING_MODEL") || DEFAULT_OPENROUTER_EMBEDDING_MODEL,
+    chatModel: Deno.env.get("OPEN_BRAIN_CHAT_MODEL") || DEFAULT_OPENROUTER_CHAT_MODEL,
+  };
+}
+
+const aiProvider = buildAiProviderConfig();
 const supabase = createClient(SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY);
 
 type ThoughtMatch = {
@@ -44,34 +109,34 @@ function thoughtUrl(id: string): string {
 }
 
 async function getEmbedding(text: string): Promise<number[]> {
-  const r = await fetch(`${OPENROUTER_BASE}/embeddings`, {
+  const r = await fetch(`${aiProvider.baseUrl}/embeddings`, {
     method: "POST",
     headers: {
-      Authorization: `Bearer ${OPENROUTER_API_KEY}`,
+      Authorization: `Bearer ${aiProvider.apiKey}`,
       "Content-Type": "application/json",
     },
     body: JSON.stringify({
-      model: "openai/text-embedding-3-small",
+      model: aiProvider.embeddingModel,
       input: text,
     }),
   });
   if (!r.ok) {
     const msg = await r.text().catch(() => "");
-    throw new Error(`OpenRouter embeddings failed: ${r.status} ${msg}`);
+    throw new Error(`${aiProvider.name} embeddings failed: ${r.status} ${msg}`);
   }
   const d = await r.json();
   return d.data[0].embedding;
 }
 
 async function extractMetadata(text: string): Promise<Record<string, unknown>> {
-  const r = await fetch(`${OPENROUTER_BASE}/chat/completions`, {
+  const r = await fetch(`${aiProvider.baseUrl}/chat/completions`, {
     method: "POST",
     headers: {
-      Authorization: `Bearer ${OPENROUTER_API_KEY}`,
+      Authorization: `Bearer ${aiProvider.apiKey}`,
       "Content-Type": "application/json",
     },
     body: JSON.stringify({
-      model: "openai/gpt-4o-mini",
+      model: aiProvider.chatModel,
       response_format: { type: "json_object" },
       messages: [
         {
diff --git a/server/test-ai-provider-config.mjs b/server/test-ai-provider-config.mjs
new file mode 100644
index 000000000..1b4124c5d
--- /dev/null
+++ b/server/test-ai-provider-config.mjs
@@ -0,0 +1,23 @@
+/**
+ * Verifies the Open Brain MCP server exposes Atlas Cloud as an optional
+ * OpenAI-compatible backend without changing OpenRouter as the default path.
+ */
+
+import assert from "node:assert/strict";
+import { readFileSync } from "node:fs";
+
+const source = readFileSync(new URL("./index.ts", import.meta.url), "utf8");
+
+assert.match(source, /OPEN_BRAIN_AI_PROVIDER/, "provider selector env is documented in code");
+assert.match(source, /atlascloud/, "Atlas Cloud selector is supported");
+assert.match(source, /ATLASCLOUD_API_KEY/, "Atlas Cloud API key env is supported");
+assert.match(source, /ATLAS_CLOUD_API_KEY/, "Atlas Cloud alias env is supported");
+assert.match(source, /https:\/\/api\.atlascloud\.ai\/v1/, "Atlas Cloud base URL is configured");
+assert.match(source, /qwen\/qwen3\.5-flash/, "Atlas Cloud chat default model is configured");
+assert.match(source, /text-embedding-3-small/, "Atlas-compatible embedding default is configured");
+assert.match(source, /openai\/gpt-4o-mini/, "OpenRouter default chat model is preserved");
+assert.match(source, /openai\/text-embedding-3-small/, "OpenRouter default embedding model is preserved");
+assert.match(source, /\$\{aiProvider\.baseUrl\}\/embeddings/, "embeddings use selected provider base URL");
+assert.match(source, /\$\{aiProvider\.baseUrl\}\/chat\/completions/, "chat uses selected provider base URL");
+
+console.log("AI provider config checks passed");