diff --git a/.env.example b/.env.example
index 48d0ee2..9333f62 100644
--- a/.env.example
+++ b/.env.example
@@ -10,10 +10,14 @@ PUBLIC_BASE_URL=http://localhost:4000
 # --- LLM providers & failover ---
 # Ordered chain of `provider:model`, tried left to right. If the first provider
 # is rate-limited, out of quota, or down, the next takes over mid-turn.
-# Providers: gmi | openrouter   (both OpenAI-compatible)
-LLM_CHAIN_SUPERVISOR=gmi:MiniMaxAI/MiniMax-M3,openrouter:nvidia/nemotron-3-ultra-550b-a55b:free
-LLM_CHAIN_AGENT=gmi:MiniMaxAI/MiniMax-M3,openrouter:nvidia/nemotron-3-ultra-550b-a55b:free
-LLM_DEFAULT_PROVIDER=gmi
+# Providers: openai | gmi | openrouter   (all OpenAI-compatible)
+LLM_CHAIN_SUPERVISOR=openai:gpt-4o-mini
+LLM_CHAIN_AGENT=openai:gpt-4o-mini
+LLM_DEFAULT_PROVIDER=openai
+
+# OpenAI
+OPENAI_API_KEY=
+OPENAI_BASE_URL=https://api.openai.com/v1
 
 # GMI Cloud — free MiniMax allocation (MiniMaxAI/MiniMax-M3, MiniMaxAI/MiniMax-M2.7)
 GMI_API_KEY=
diff --git a/src/config/index.js b/src/config/index.js
index 3459885..5860e08 100644
--- a/src/config/index.js
+++ b/src/config/index.js
@@ -18,13 +18,17 @@ const config = {
 
   llm: {
     // Ordered failover chain of `provider:model` specs, tried left to right.
-    supervisorChain: list(process.env.LLM_CHAIN_SUPERVISOR, ['gmi:MiniMaxAI/MiniMax-M3']),
-    agentChain: list(process.env.LLM_CHAIN_AGENT, ['gmi:MiniMaxAI/MiniMax-M3']),
+    supervisorChain: list(process.env.LLM_CHAIN_SUPERVISOR, ['openai:gpt-4o-mini']),
+    agentChain: list(process.env.LLM_CHAIN_AGENT, ['openai:gpt-4o-mini']),
 
     // Every provider is OpenAI-compatible; they differ only by base URL + key.
     // A provider with no key is skipped when a chain references it.
-    defaultProvider: process.env.LLM_DEFAULT_PROVIDER || 'gmi',
+    defaultProvider: process.env.LLM_DEFAULT_PROVIDER || 'openai',
     providers: {
+      openai: {
+        apiKey: process.env.OPENAI_API_KEY,
+        baseUrl: process.env.OPENAI_BASE_URL || 'https://api.openai.com/v1',
+      },
       gmi: {
         apiKey: process.env.GMI_API_KEY,
         baseUrl: process.env.GMI_BASE_URL || 'https://api.gmi-serving.com/v1',
diff --git a/src/server.js b/src/server.js
index 9a613c0..7feb9ac 100644
--- a/src/server.js
+++ b/src/server.js
@@ -57,7 +57,7 @@ app.use((err, _req, res, _next) => {
 async function start() {
   const configured = Object.entries(config.llm.providers).filter(([, p]) => p.apiKey).map(([n]) => n);
   if (!configured.length) {
-    logger.error('No provider API key set (GMI_API_KEY / OPENROUTER_API_KEY) — the assistant cannot run. Add one to .env.');
+    logger.error('No provider API key set (OPENAI_API_KEY / GMI_API_KEY / OPENROUTER_API_KEY) — the assistant cannot run. Add one to .env.');
     process.exit(1);
   }
   logger.info(`   providers : ${configured.join(', ')}`);