How to Fix Broken Markdown Formatting in Claude Messages from n8n

Run your workflow manually and click on each node in sequence to inspect the output. Start from the Claude or AI Agent node and trace the response through every downstream node. In the output panel, switch between Table view and JSON view to see how the markdown text is stored. In JSON view, look for double-escaped characters: \\n instead of \n for newlines, \\* instead of * for asterisks, or missing backticks. The node where the formatting first breaks is where you need to add a fix. Common culprits include Set nodes that re-assign the text value, Function/Code nodes that call JSON.stringify without handling special characters, and HTTP Request or Webhook Response nodes that encode the output.

Insert a Code node immediately after the node where formatting breaks. This node unescapes double-escaped characters that commonly corrupt markdown. Set it to Run Once for All Items. The code processes the Claude response text and replaces double-escaped newlines, tabs, asterisks, and backticks with their single-escaped equivalents. It also normalizes line endings and preserves code block indentation. Place this node before any output node like Respond to Webhook, Slack, or Email.

1// Code node: Fix Markdown Escaping
2// Mode: Run Once for All Items
3
4const items = $input.all();
5
6for (const item of items) {
7  let text = item.json.output || item.json.text || item.json.message || '';
8
9  // Fix double-escaped newlines
10  text = text.replace(/\\n/g, '\n');
11
12  // Fix double-escaped tabs
13  text = text.replace(/\\t/g, '\t');
14
15  // Fix escaped asterisks (broken bold/italic)
16  text = text.replace(/\\\*/g, '*');
17
18  // Fix escaped backticks (broken code blocks)
19  text = text.replace(/\\`/g, '`');
20
21  // Fix escaped brackets (broken links)
22  text = text.replace(/\\\[/g, '[');
23  text = text.replace(/\\\]/g, ']');
24
25  // Normalize line endings
26  text = text.replace(/\r\n/g, '\n');
27
28  item.json.formattedOutput = text;
29}
30
31return items;

If your workflow returns the Claude response via Respond to Webhook, incorrect content type settings cause markdown to render as plain text or get HTML-encoded. In the Respond to Webhook node, set the Response Content Type to text/plain if you want the raw markdown, or text/html if you want to render it. For API consumers that parse markdown (like chat UIs), use application/json and wrap the markdown in a JSON field. If using text/html, you need to convert markdown to HTML first using a Code node with a simple markdown-to-HTML converter. Do NOT use application/json content type if you are putting raw text directly in the response body, as this will add extra JSON escaping.

1// For Respond to Webhook sending raw markdown:
2// Response Content Type: text/plain
3// Response Body: {{ $json.formattedOutput }}
4
5// For API consumers expecting JSON:
6// Response Content Type: application/json
7// Response Body:
8{
9  "response": "{{ $json.formattedOutput }}",
10  "status": "success"
11}

Set nodes and Merge nodes can strip whitespace from text values, destroying code block indentation. When you need to pass Claude's response through a Set node, use the expression {{ $json.output }} directly rather than assigning it to a new field and then referencing it. If you must use a Set node, set the value type to String and use the expression editor rather than the fixed value input. For Merge nodes, use the Combine mode and ensure the text field from the Claude branch is the one preserved, not overwritten by the other branch. Add a test Code node after the Merge that validates the code blocks still have their original indentation.

1// Code node: Validate code block preservation
2const items = $input.all();
3
4for (const item of items) {
5  const text = item.json.formattedOutput || item.json.output || '';
6
7  // Check if code blocks still have indentation
8  const codeBlockRegex = /```[\s\S]*?```/g;
9  const codeBlocks = text.match(codeBlockRegex) || [];
10
11  for (const block of codeBlocks) {
12    const lines = block.split('\n');
13    const hasIndentation = lines.some(line => line.startsWith('  ') || line.startsWith('\t'));
14    if (!hasIndentation && lines.length > 2) {
15      console.log('WARNING: Code block may have lost indentation');
16    }
17  }
18}
19
20return items;

Slack and Email nodes have their own markdown interpretation. Slack uses mrkdwn format which differs from standard markdown: bold is *text* not **text**, and code blocks use triple backticks without a language specifier working differently. Email nodes may strip markdown entirely or render it as plain text. For Slack, add a Code node that converts standard markdown to Slack mrkdwn before the Slack node. For Email, convert markdown to HTML. If you need both, branch the workflow and apply different formatting transformations for each destination.

1// Code node: Convert standard markdown to Slack mrkdwn
2const items = $input.all();
3
4for (const item of items) {
5  let text = item.json.formattedOutput || '';
6
7  // Convert standard bold **text** to Slack bold *text*
8  text = text.replace(/\*\*(.+?)\*\*/g, '*$1*');
9
10  // Convert headers ## to bold
11  text = text.replace(/^#{1,6}\s+(.+)$/gm, '*$1*');
12
13  // Convert standard italic _text_ to Slack italic _text_ (same)
14  // Convert links [text](url) to Slack <url|text>
15  text = text.replace(/\[(.+?)\]\((.+?)\)/g, '<$2|$1>');
16
17  item.json.slackMessage = text;
18}
19
20return items;