Claude produces better-structured markdown than any other LLM — consistent headings, clean tables, properly tagged code blocks. The only problem is getting that structure into a PDF without losing it. MarkdownTools does that in 30 seconds.
Claude uses heading levels consistently, writes tables with proper alignment, and tags code blocks with the correct language identifier — even without explicit formatting instructions. This is not typical behavior for large language models. Most require careful prompt engineering to produce clean markdown. Claude does it by default.
The result: Claude outputs paste into MarkdownTools and render correctly on the first try. You are not fixing heading levels or cleaning up broken table syntax. You are going straight from Claude to PDF in 30 seconds.
Heading hierarchy
Consistent # → ## → ###
Code block tags
Correct language IDs
Table alignment
Proper column syntax
Typical Claude output
# Architecture Decision Record ## Context The **authentication service** needs to support both OAuth and API key flows. | Flow | Latency | Use case | |------|---------|----------| | OAuth | ~200ms | Interactive | | API key | ~10ms | Programmatic | ## Decision Implement both via middleware: ```typescript type AuthFlow = 'oauth' | 'apikey' ```
Four steps. Works for inline responses, artifacts, and extended thinking outputs.
Select the Claude response text — click anywhere in the message, then Cmd+A selects the full message. Copy it. If you are working with a Claude artifact, use the copy icon in the artifact toolbar. About 5 seconds.
Open MarkdownTools and paste into the editor. Claude's heading hierarchy, tables, and code blocks all render correctly in the preview on the right. Extended thinking outputs and multi-section documents are handled without any cleanup. 5 seconds.
Choose Clean for technical documentation or developer-facing content, Classic for reports and summaries, or Elegant for client-facing proposals and executive documents. The live preview updates immediately. 10 seconds.
Click the PDF button. A professionally formatted PDF is generated server-side using Chromium and downloads to your device in under a second. Your content is never stored. 5 seconds.
Claude does not need much help, but these additions consistently improve the result.
"Write this as a [document type]: technical spec / client proposal / architecture decision record."
Claude adopts the structural conventions of the document type without further instruction. The most effective single addition to any prompt where the output will become a PDF.
"Present any comparisons or structured data as markdown tables."
Claude sometimes defaults to bullet lists for data that belongs in a table. A table renders significantly better in PDF and is faster to read at a glance.
"Use # for the document title, ## for main sections, and ### for subsections. No deeper nesting."
Enforces two heading levels, which is the professional standard for most documents. Rarely needed with Claude — but useful for long documents where heading depth can drift.
"Tag all code examples with the correct language: python, typescript, sql, bash, etc."
Claude usually does this automatically. Adding it explicitly is useful when you are asking about unfamiliar languages where Claude might guess or omit the tag.
"Provide a CLAUDE.md-style project context section at the top of the document."
Useful for technical documentation that will be used with Claude Code later. The output is already in the right format to copy into a project CLAUDE.md — or to convert to PDF for human readers.
"End each major section with a one-sentence summary of the key conclusion."
Produces documents that are easier to scan in PDF format. Readers of printed or PDF documents scan before they read — a section summary helps them decide what to read closely.
Claude-specific output types that come up regularly.
Claude artifacts
Use the copy icon at the top of the artifact panel — not Cmd+A in the artifact body. The copy icon captures the raw markdown. Artifact markdown is typically the cleanest output Claude produces.
Extended thinking outputs
The "Thinking" preamble is usually not for sharing. Select and delete it in the MarkdownTools editor before exporting. The main response body converts cleanly — extended thinking outputs are some of the most structured content Claude produces.
CLAUDE.md project files
Open the file in any editor, copy the full contents, paste into MarkdownTools. The structured sections that make a good CLAUDE.md — overview, commands, architecture — render as a clean reference document for team members.
Multi-section responses with dividers
Claude sometimes uses --- horizontal rules to separate major sections. These render as visual dividers in the PDF — a useful structural element that most other LLMs do not use consistently.
Responses with nested blockquotes
Claude uses blockquotes for source material, callouts, and key takeaways. These render correctly in MarkdownTools — indented, with a left border — and look professional in the PDF.
Wide tables with many columns
The one case where Claude output sometimes needs fixing. Ask Claude to split a wide table into two narrower ones grouped by category, or to present the same data as a definition list. Both alternatives render better in standard page widths.
Things that come up with Claude outputs specifically.
Long document breaks pagination badly
Why: Extended thinking outputs or long architecture documents sometimes have page breaks mid-code-block or mid-table.
Fix: Split the document into logical sections and export each as a separate PDF. Or reduce font size by switching to the Clean theme, which is the most compact of the three.
Code block spans multiple pages awkwardly
Why: A long code example hits a page break in the middle, which is hard to read.
Fix: In the editor, break the code block into smaller examples with explanatory text between them. This is usually better documentation anyway — short, annotated examples are easier to follow than one monolithic block.
Table overflows the page margins
Why: Claude generated a table with more columns than fit in the page width.
Fix: Ask Claude: "split that table by grouping the left three columns together and the right columns separately." Or ask it to present the same data as a definition list, which flows naturally within page margins.
"Thinking" section appears in the PDF
Why: The extended thinking preamble was included when you copied the response.
Fix: In the MarkdownTools editor, select and delete everything before the first # heading of the main response. The Thinking section does not start with a heading, so it is easy to identify and remove.
Markdown symbols showing as raw text
Why: You copied the rendered Claude response text instead of the raw markdown.
Fix: For artifacts: use the copy icon in the artifact toolbar. For inline responses: select the text carefully — Claude renders markdown in the UI, so copying by selecting will sometimes give you rendered text. If you see ## showing literally, click the "copy" icon below the message instead.
No syntax highlighting on code blocks
Why: Claude used an unusual or abbreviated language tag — sometimes happens with niche languages.
Fix: In the editor, change the language tag after the opening triple backtick to a recognized identifier: typescript, python, sql, bash, yaml, json, rust, go. The preview updates immediately.
Anywhere you would normally wrestle a Claude response into a word processor and lose an hour reformatting it.
Ask Claude to generate API reference documentation, architecture decision records, or setup guides. Claude's consistent heading structure and code block formatting produces technical docs that render perfectly — syntax-highlighted code, proper table alignment, and a heading hierarchy that translates directly into a scannable PDF.
CLAUDE.md files — used by Claude Code to provide project context — are valid markdown. If you want a shareable, printable version of your project's CLAUDE.md for onboarding or handoffs, paste it into MarkdownTools and export it as a formatted PDF in one step.
Claude's extended thinking mode produces long, deeply reasoned documents. These outputs are almost always well-structured markdown. Paste a full thinking output into MarkdownTools and the entire document — headings, subheadings, code examples, and tables — renders correctly without any manual cleanup.
Draft a strategy memo, competitive analysis, or project proposal with Claude using structured sections. The Elegant theme converts that markdown output into a document that looks like it came from a design-conscious team, not a text editor. Send it directly to the client.
Claude handles multi-document synthesis well and produces structured summaries with clear section breaks. Convert those summaries to PDF for offline reading, sharing with colleagues who do not use Claude, or archiving research threads in a format that does not require a browser to open.
For larger documentation projects, use Claude to generate individual sections or modules, then convert each to PDF separately. Claude's consistent formatting means each file uses the same heading conventions, so the final PDF package looks cohesive even if generated across multiple conversations.
Claude is trained with a strong preference for structured, well-organized output. In practice this means Claude uses heading levels consistently (# for titles, ## for sections, ### for subsections), nests bullet points correctly, writes tables with proper column alignment, and uses fenced code blocks with accurate language tags. It does this even without explicit formatting instructions in the prompt.
Other models are more erratic — they will sometimes produce heading soup (too many H2s, no H1), inconsistent code block language tags, or bullet lists where a table would be more appropriate. Claude defaults toward more document-like structure, which is why its output converts to PDF so cleanly.
The practical result: when you paste a Claude response into MarkdownTools, it almost always previews correctly on the first try. You rarely need to fix heading levels, clean up broken tables, or manually add language tags to code blocks before exporting.
Claude artifacts — the separate panel that appears for longer, structured outputs — contain clean markdown that is ready to convert. Click the copy icon at the top of the artifact panel to copy the raw markdown content. Do not copy from the rendered preview, as that copies formatted text with some markdown symbols stripped.
Artifacts are particularly clean because Claude tends to write them with the intention of sharing, so it naturally uses better heading structure and fewer conversational asides than inline responses. Paste the copied artifact content directly into MarkdownTools — it typically needs zero editing before export.
If you are using the Claude API and storing artifact content programmatically, the same logic applies: the raw markdown string that Claude returns is exactly what you paste into MarkdownTools. No preprocessing needed.
Claude's extended thinking mode produces outputs that can run to thousands of words across many sections. These are some of the most document-like outputs any LLM produces — they have clear thesis statements, structured arguments, and consistent section formatting.
The MarkdownTools editor handles documents up to 10,000 lines. For a typical extended thinking output of 3,000 to 6,000 words, PDF generation completes in under two seconds. The heading hierarchy that Claude establishes carries through into the PDF, so long documents remain navigable.
One practical note: extended thinking outputs sometimes include a "Thinking" preamble before the final response. That section is usually not meant for sharing. In the MarkdownTools editor, simply select and delete the preamble before exporting. The rest of the document converts cleanly.
Claude needs less prompt engineering for PDF-ready output than other models, but a few additions reliably improve the result. The most impactful: tell Claude the document type upfront. "Write this as a technical design document" or "format this as a client proposal" causes Claude to adopt appropriate structural conventions without further instruction.
For data-heavy content, explicitly ask for tables: "present the comparison as a markdown table." Claude will sometimes default to bullet lists for structured data, especially in shorter responses. A table renders significantly better in PDF and is faster to read.
If you need a document with a predictable structure — for example, a weekly status report you generate repeatedly — give Claude a heading template in the system prompt. Claude follows heading templates very reliably, and the consistent structure makes each export visually identical.
Everything you need to know.
Yes. Extended thinking outputs are long, well-structured markdown documents that convert cleanly. Paste the full output into MarkdownTools — the editor handles documents up to 10,000 lines. If the output includes a "Thinking" preamble you do not want to share, delete it in the editor before exporting. The rest of the document will convert without any cleanup needed.
Paste it. Pick a theme. Download. No signup, no formatting work, no cost.