diff --git a/src/App.jsx b/src/App.jsx
index a2d3405..1a244ea 100644
--- a/src/App.jsx
+++ b/src/App.jsx
@@ -1993,6 +1993,7 @@ async function createNewChat() {
             onStreamOutputChange={setStreamOutput}
             onTranscriptionModelChange={setTranscriptionModel}
             onVisionModelChange={setVisionModel}
+            onWorkflowRouterModelChange={setWorkflowRouterModel}
             searxEngines={searxEngines}
             searxUrl={searxUrl}
             setSearxEngines={setSearxEngines}
diff --git a/src/GeneralSettings.jsx b/src/GeneralSettings.jsx
index 93eea15..5237457 100644
--- a/src/GeneralSettings.jsx
+++ b/src/GeneralSettings.jsx
@@ -15,6 +15,7 @@ const ENRICHMENT_MODEL_KEY = 'enrichmentModel';
 const MODEL_KEY = 'chatModel';
 const VISION_MODEL_KEY = 'visionModel';
 const TRANSCRIPTION_MODEL_KEY = 'transcriptionModel';
+const WORKFLOW_ROUTER_MODEL_KEY = 'workflowRouterModel';
 const AUTO_DEEP_ENRICHMENT_KEY = 'autoDeepEnrichment';
 const DEFAULT_AUDIO_INPUT_DEVICE_ID = '';
 const DEFAULT_AUDIO_INPUT_LANGUAGE = '';
@@ -74,6 +75,7 @@ export default function GeneralSettings({
   panel,
   onModelChange,
   onVisionModelChange,
+  onWorkflowRouterModelChange,
   onTranscriptionModelChange,
   onLibrariesPurged,
   onAudioInputDeviceChange,
@@ -91,6 +93,7 @@ export default function GeneralSettings({
   const [selectedModel, setSelectedModel] = useState('');
   const [visionModel, setVisionModel] = useState('');
   const [transcriptionModel, setTranscriptionModel] = useState(DEFAULT_TRANSCRIPTION_MODEL);
+  const [workflowRouterModel, setWorkflowRouterModel] = useState('');
   const [autoDeepEnrichment, setAutoDeepEnrichment] = useState(true);
   const [audioInputDeviceId, setAudioInputDeviceId] = useState(DEFAULT_AUDIO_INPUT_DEVICE_ID);
   const [audioInputLanguage, setAudioInputLanguage] = useState(DEFAULT_AUDIO_INPUT_LANGUAGE);
@@ -129,6 +132,7 @@ export default function GeneralSettings({
       setSelectedModel(settings.chatModel || '');
       setVisionModel(settings.visionModel || settings.chatModel || '');
       setTranscriptionModel(settings.transcriptionModel || DEFAULT_TRANSCRIPTION_MODEL);
+      setWorkflowRouterModel(settings.workflowRouterModel || '');
       setAutoDeepEnrichment(settings.autoDeepEnrichment !== false);
       setAudioInputDeviceId(
         typeof settings.audioInputDeviceId === 'string'
@@ -419,6 +423,13 @@ export default function GeneralSettings({
     }
   };
 
+  const handleWorkflowRouterModelChange = (event) => {
+    const nextModel = event.target.value;
+    setWorkflowRouterModel(nextModel);
+    desktopApi.setSetting(WORKFLOW_ROUTER_MODEL_KEY, nextModel);
+    onWorkflowRouterModelChange?.(nextModel);
+  };
+
   const handleAutoDeepEnrichmentToggle = () => {
     const nextValue = !autoDeepEnrichment;
     setAutoDeepEnrichment(nextValue);
@@ -605,6 +616,7 @@ export default function GeneralSettings({
   const rerankingModelOptions = buildSelectOptions(rerankingModels, rerankModel, 'saved model unavailable', showMissingModelLabel);
   const enrichmentModelOptions = buildSelectOptions(chatModels, enrichmentModel, 'saved model unavailable', showMissingModelLabel);
   const transcriptionModelOptions = buildSelectOptions(whisperModels, transcriptionModel, 'saved model unavailable', showMissingModelLabel);
+  const workflowRouterModelOptions = buildSelectOptions(chatModels, workflowRouterModel, 'saved model unavailable', showMissingModelLabel);
 
   if (panel === 'AI Models') {
     return (
@@ -623,6 +635,16 @@ export default function GeneralSettings({
             Heimgeist uses this model for normal text chat.
           </p>
         </div>
+        <div className="setting-section">
+          <h3>Workflow Router Model</h3>
+          <select className="select" value={workflowRouterModel} onChange={handleWorkflowRouterModelChange}>
+            <option value="">Use chat model</option>
+            {workflowRouterModelOptions.map(model => <option key={model.value} value={model.value}>{model.label}</option>)}
+          </select>
+          <p className="setting-description">
+            Automatic workflow selection uses this local Ollama model and falls back to the chat model when unavailable.
+          </p>
+        </div>
         <div className="setting-section">
           <h3>Vision Model</h3>
           <select
diff --git a/src/SettingsPanels.jsx b/src/SettingsPanels.jsx
index 09c76de..2dcc3e0 100644
--- a/src/SettingsPanels.jsx
+++ b/src/SettingsPanels.jsx
@@ -31,6 +31,7 @@ export function SettingsPanel({
   onStreamOutputChange,
   onTranscriptionModelChange,
   onVisionModelChange,
+  onWorkflowRouterModelChange,
   searxEngines,
   searxUrl,
   setSearxEngines,
@@ -54,6 +55,7 @@ export function SettingsPanel({
           panel="AI Models"
           onModelChange={onModelChange}
           onVisionModelChange={onVisionModelChange}
+          onWorkflowRouterModelChange={onWorkflowRouterModelChange}
           onTranscriptionModelChange={onTranscriptionModelChange}
           onLibrariesPurged={onLibrariesPurged}
         />
diff --git a/src/desktop/desktopApi.js b/src/desktop/desktopApi.js
index a75e6e9..1120f12 100644
--- a/src/desktop/desktopApi.js
+++ b/src/desktop/desktopApi.js
@@ -6,6 +6,7 @@ const DEFAULT_SETTINGS = Object.freeze({
   embedModel: 'nomic-embed-text:latest',
   rerankModel: 'nomic-embed-text:latest',
   transcriptionModel: 'base',
+  workflowRouterModel: '',
   autoDeepEnrichment: true,
   colorScheme: 'Default',
   uiScale: 1,