[BUG] Claude VSCode Extension: Chat output splits every word or character into a new line

Found 3 possible duplicate issues:

1. https://github.com/anthropics/claude-code/issues/12290
2. https://github.com/anthropics/claude-code/issues/12157
3. https://github.com/anthropics/claude-code/issues/6827

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

Root Cause: VSCode Markdown Rendering Bug

Your chat output is being treated as Markdown list items instead of plain text.

Why This Happens

VSCode's webview renderer sees each line break and interprets it as:

I'm
doing
well
,
thanks
!

Instead of:

I'm doing well, thanks!

Root cause: Claude Code VSCode extension is inserting \n after every word/token in streaming response.

Immediate Workaround

Option 1: Use Terminal Instead

# Don't use VSCode extension, use standalone CLI
npm install -g @anthropic-ai/claude-code
claude
# Output renders correctly in terminal

Option 2: Copy/Paste Output

- Click "Copy" button on Claude's response
- Paste into text editor
- Text will be properly formatted (single lines)

Option 3: Downgrade Extension

# In VSCode:
# Extensions → Claude Code → Gear icon → Install Another Version
# Select 2.0.49 or earlier (before regression)

Technical Analysis

Suspect code location (in VSCode extension):

// Broken streaming handler:
response.on('data', (chunk) => {
  const token = chunk.toString();
  webview.postMessage({
    type: 'append',
    content: token + '\n'  // ← BUG: Extra newline
  });
});

// Should be:
response.on('data', (chunk) => {
  const token = chunk.toString();
  webview.postMessage({
    type: 'append',
    content: token  // No newline unless token ends with \n
  });
});

Proper Fix (For Anthropic Team)

In webview message handler:

// src/webview/ChatRenderer.tsx (or similar)

handleAppendMessage(message: AppendMessage) {
  const { content } = message;
  
  // Buffer tokens until sentence boundary
  this.textBuffer += content;
  
  // Only render complete sentences
  if (content.match(/[.!?]\s*$/)) {
    this.renderText(this.textBuffer);
    this.textBuffer = '';
  }
}

Or simpler - remove extra newlines:

handleAppendMessage(message: AppendMessage) {
  // Strip extra newlines from streamed tokens
  const cleaned = message.content.replace(/\n+/g, ' ');
  this.appendToChat(cleaned);
}

Verification

After fix:

// Test case:
const testInput = "I'm\ndoing\nwell\n,\nthanks\n!";
const output = cleanStreamedText(testInput);
console.assert(output === "I'm doing well, thanks!");

Why This is a Regression

v2.0.49: Worked correctly
v2.0.50+: Started splitting every word

Likely change: Streaming implementation was modified to improve responsiveness but introduced newline bug.

Related Issues

This might be the same root cause as:

• Terminal text rendering issues
• Copy/paste formatting problems
• Markdown parsing errors in chat

Suspect: Unified text rendering pipeline was changed across all interfaces.

---

Workaround: Use terminal CLI instead of VSCode extension until fixed.
Priority: P1 - Breaks primary use case (VSCode chat)

@JuanCS-Dev it also happened to Claude Code CLI

this is happening with me in cli and the extensions both. i tried downgrading it but then it was not able to receive any response. i tried downgrading to all from latest to 2.0.35.

please fix this asap

Hi, It is not working on both claude code cli and claude code vs code extension. Both of them are basically unusable now due to the new lines after each word generation. This is a critical issue, please fix it

Happened to me in the CLI, but only on Mac. On Linux, it work fine

just tested it, it working for me in version 2.0.54

After updating to version 2.0.55, the issue no longer occurs.

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.