diff --git a/chat/public/builder.js b/chat/public/builder.js
index b7696f1..b0e6417 100644
--- a/chat/public/builder.js
+++ b/chat/public/builder.js
@@ -1896,6 +1896,17 @@ function hideLoadingIndicator() {
 window.hideLoadingIndicator = hideLoadingIndicator;
 
 function renderMessages(session) {
+  // Helper function to strip markdown formatting from user input
+  function stripMarkdownFormatting(text) {
+    if (!text || typeof text !== 'string') return text;
+    let cleaned = text;
+    cleaned = cleaned.replace(/\*\*([^*]+)\*\*/g, '$1');
+    cleaned = cleaned.replace(/\*\*/g, '');
+    cleaned = cleaned.replace(/^##\s+/gm, '');
+    cleaned = cleaned.replace(/##/g, '');
+    return cleaned;
+  }
+
   // Don't clear the loading indicator if it exists
   const loadingIndicator = document.getElementById('loading-indicator');
   // Store reference before clearing, and remove animation class to prevent bounce on re-render
@@ -1960,7 +1971,8 @@ function renderMessages(session) {
     `;
     const userBody = document.createElement('div');
     userBody.className = 'body';
-    userBody.appendChild(renderContentWithTodos(msg.displayContent || msg.content || ''));
+    const contentToRender = isPlanMessage ? stripMarkdownFormatting(msg.displayContent || msg.content || '') : (msg.displayContent || msg.content || '');
+    userBody.appendChild(renderContentWithTodos(contentToRender));
     userCard.appendChild(userMeta);
     userCard.appendChild(userBody);
     // Attachments
@@ -3190,7 +3202,18 @@ function streamMessage(sessionId, messageId) {
     const message = session?.messages.find(m => m.id === messageId);
     
     if (message && message.status === 'queued') {
-      // Server restart was signaled, poll for reconnection
+      // Check if this is a plan message (don't set error for plan messages, just log)
+      const isPlanMessage = message.phase === 'plan' || message.cli !== 'opencode';
+      
+      if (isPlanMessage) {
+        // For plan messages, just log and don't modify the message
+        console.log('[SSE] Plan message encountered server restart, preserving original content', { messageId, sessionId });
+        hideLoadingIndicator();
+        setStatus('Service temporarily unavailable. Please try again in a moment.');
+        return;
+      }
+      
+      // For OpenCode messages, poll for reconnection
       console.log('Server restart detected, attempting reconnection...', { messageId, sessionId });
       let reconnectAttempts = 0;
       const maxReconnectAttempts = 30; // 30 seconds max
diff --git a/chat/server.js b/chat/server.js
index 6d72c8e..a90349a 100644
--- a/chat/server.js
+++ b/chat/server.js
@@ -9038,9 +9038,22 @@ async function handlePlanMessage(req, res, userId) {
   const session = getSession(sessionId, userId);
   if (!session) return sendJson(res, 404, { error: 'Session not found' });
 
+  // Helper function to strip markdown formatting from display content
+  function stripMarkdownFromDisplay(text) {
+    if (!text || typeof text !== 'string') return text;
+    let cleaned = text;
+    cleaned = cleaned.replace(/\*\*([^*]+)\*\*/g, '$1');
+    cleaned = cleaned.replace(/\*\*/g, '');
+    cleaned = cleaned.replace(/^##\s+/gm, '');
+    cleaned = cleaned.replace(/##/g, '');
+    return cleaned;
+  }
+
   // Sanitize user input to prevent prompt injection
   const content = sanitizePromptInput(body.content || '');
-  const displayContent = typeof body.displayContent === 'string' && body.displayContent.trim() ? sanitizePromptInput(body.displayContent) : content;
+  const displayContent = typeof body.displayContent === 'string' && body.displayContent.trim() ? 
+    stripMarkdownFromDisplay(sanitizePromptInput(body.displayContent)) : 
+    stripMarkdownFromDisplay(content);
   if (!content) return sendJson(res, 400, { error: 'Message is required' });
   const userPlan = resolveUserPlan(session.userId);
   const allowance = canConsumeTokens(session.userId, userPlan, estimateTokensFromText(content) + TOKEN_ESTIMATION_BUFFER);
@@ -10293,7 +10306,7 @@ function shouldFallbackCliError(err, message) {
     // Rate limiting - these require fallback
     /error:.*rate limit exceeded/i,
     /error:.*too many requests/i,
-    /error:.*rate limited/i,
+    /error:.*rate limited/gi,
     // Size limits - clear model constraints
     /error:.*request too large/i,
     /error:.*token limit exceeded/i,