diff --git a/README.md b/README.md
index 78d7dcaa..4dc4eb5c 100644
--- a/README.md
+++ b/README.md
@@ -102,10 +102,10 @@ FIRECRAWL_KEY="your_firecrawl_key"
 OPENAI_KEY="your_openai_key"
 ```
 
-To use local LLM, comment out `OPENAI_KEY` and instead uncomment `OPENAI_ENDPOINT` and `OPENAI_MODEL`:
+To use a local LLM, uncomment `OPENAI_ENDPOINT` and `CUSTOM_MODEL` (you do not need to set `OPENAI_KEY` when using a local endpoint):
 
-- Set `OPENAI_ENDPOINT` to the address of your local server (eg."http://localhost:1234/v1")
-- Set `OPENAI_MODEL` to the name of the model loaded in your local server.
+- Set `OPENAI_ENDPOINT` to the address of your local server (eg."http://localhost:11434/v1")
+- Set `CUSTOM_MODEL` to the name of the model loaded in your local server.
 
 ### Docker
 
diff --git a/src/ai/providers.ts b/src/ai/providers.ts
index ce305aa3..f4290aa3 100644
--- a/src/ai/providers.ts
+++ b/src/ai/providers.ts
@@ -10,12 +10,15 @@ import { getEncoding } from 'js-tiktoken';
 import { RecursiveCharacterTextSplitter } from './text-splitter';
 
 // Providers
-const openai = process.env.OPENAI_KEY
-  ? createOpenAI({
-      apiKey: process.env.OPENAI_KEY,
-      baseURL: process.env.OPENAI_ENDPOINT || 'https://api.openai.com/v1',
-    })
-  : undefined;
+// Create openai client when OPENAI_KEY is set, or when CUSTOM_MODEL is configured
+// (local/OpenAI-compatible endpoints don't require a real API key)
+const openai =
+  process.env.OPENAI_KEY || process.env.CUSTOM_MODEL
+    ? createOpenAI({
+        apiKey: process.env.OPENAI_KEY || 'placeholder',
+        baseURL: process.env.OPENAI_ENDPOINT || 'https://api.openai.com/v1',
+      })
+    : undefined;
 
 const fireworks = process.env.FIREWORKS_KEY
   ? createFireworks({