Formatting & Response Style
Claude formats replies with Markdown — headings, lists, tables, and code blocks render in the chat. You control length and structure by saying what you want in the prompt, or by refining after the first reply.
Markdown rendering
Claude writes in Markdown, and the chat interface renders it automatically. You see formatted text, not raw symbols — though you can copy the underlying Markdown if needed.
| Element | What it does | Syntax example |
|---|---|---|
| Headers | Structure long answers into sections | ## Section title |
| Bullets and numbers | Lists for steps, items, or options | - Item one |
| Tables | Side-by-side comparisons and data | | Col A | Col B | |
| Code blocks | Syntax-highlighted code with a copy button | ```python |
Prerequisites
- Node.js 18 or later
- A code editor (VS Code recommended)
- Git installed
Quick comparison
| Tool | Speed | Best for |
|---|---|---|
| npm | Fast | Most projects |
| yarn | Fast | Monorepos |
Install command
Shorter or longer responses
Claude defaults to a balanced length — enough detail to be useful without being overwhelming. Shift that in either direction with a few words in your prompt, or refine after the first reply.
| Goal | Example phrase |
|---|---|
| Shorter | "Keep it under 100 words." / "Give me a one-paragraph summary." |
| Brief | "Answer in 2–3 bullet points." / "TL;DR only." |
| Standard | "Explain this clearly." (no length constraint) |
| Longer | "Go into detail." / "Explain step by step with examples." |
| Long form | "Cover each major section in detail." |
| Adjust after | "Make that shorter." / "Expand on the second point." |
What DNS does
DNS maps domain names to IP addresses…
How a lookup works
- Browser checks local cache
- Queries recursive resolver
- Walks the DNS hierarchy…
+ examples, record types, troubleshooting
Requesting specific formats
Tell Claude how you want the output structured. The right format depends on what you plan to do with the answer — paste into code, scan quickly, or read as a document.
| Format | Best for | Example prompt |
|---|---|---|
| JSON | API responses, config files, structured data for code | "Return only valid JSON, no explanation." |
| Bullet list | Action items, short lists, quick scanning | "List the top 5 risks as bullet points." |
| Numbered steps | Tutorials, procedures, ordered instructions | "Give me numbered steps to deploy." |
| Table | Comparisons, feature matrices, side-by-side data | "Compare these in a table with columns for price, speed, and ease." |
| Prose | Essays, emails, narratives, reports | "Write this as a professional paragraph, not a list." |
| Code only | When you just want the snippet, no commentary | "Return only the Python code, no explanation." |
Combining format and length
You can combine constraints in a single prompt. The more specific you are, the fewer follow-up messages you need.
Strong combined prompt
“Summarize this article in exactly 5 bullet points. Each bullet should be one sentence. Use plain language for a non-technical audience.”
| Vague | Specific |
|---|---|
| “Tell me about APIs.” | “Explain REST APIs in 3 bullet points for a beginner. Keep it under 80 words.” |
Before you continue
- Claude renders Markdown — headers, bullets, tables, and code blocks display formatted.
- Control length with phrases like “under 100 words” or “go into detail.”
- Request specific formats: JSON, lists, tables, prose, or code only.
- State format, length, and audience in one prompt to cut down on follow-ups.
What's Next
Claude Chat vs. Other Products — when to use Chat, Code, Cowork, Skills, or Plugins.