Why not just ask Claude to make the PDF?

Often you should. Claude creates PDF, Word, Excel and PowerPoint files on its own, and for a document you need once, that is the shortest path and nothing here beats it.

The answer changes when you make documents repeatedly. Then three things start to cost you, and none of them is whether a PDF comes out the other end.

The three differences that matter

1. You edit the document, not the prompt

Unplain has an editor. Click a paragraph, change the word, done. The change is exactly what you typed, it happens immediately, and it costs nothing.

Changing a sentence in a generated file means describing the change and having the file made again. That is a round trip, it spends tokens, and regeneration is not surgical: the model can adjust a heading, a spacing choice or a sentence you were happy with, because it is producing the document afresh rather than amending the one you had. On the fourth revision of a long report, that difference stops being theoretical.

2. Your brand is stored, not re-described

A Brand Theme holds your colors, fonts and logo. Every document you make afterwards uses it without anyone restating anything, so the tenth report matches the first because it is the same stored record rather than a fresh interpretation of the same sentence.

Telling an assistant "use our brand colors" works, once. Across a series it drifts, because each conversation reads that instruction independently. Nobody notices in a single document. Everybody notices when four reports sit side by side and the accent is slightly different in each.

3. The document still exists next month

In Unplain a document is a saved project in your account. Open it in six months, change the two numbers that went stale, export it again. The work is an asset you keep.

Files Claude creates stay available for download within the conversation that produced them. That is exactly right for a one-off and awkward for anything recurring, because next quarter's version starts from a prompt rather than from last quarter's document.

Claude's file creation details verified August 2026 against Anthropic's own documentation. Capabilities change, so check theirs if the difference matters to your decision.

Capability comparison

Dimension Unplain Claude on its own
Makes a PDF Yes, and only PDF. Yes, plus DOCX, XLSX and PPTX.
Writes the content No. You bring it, or an assistant does. Yes. That is the point of it.
Changing one paragraph Click it and type. The change is exactly what you typed, it is instant, and it costs nothing. Describe the change and the file is generated again. Costs tokens and a round trip, and the model may adjust things you did not mention.
Brand consistency Stored. A Brand Theme holds colors, fonts and logo and applies identically every time (1 on Free, up to 25 on Pro). Described. You restate the brand each conversation, and the model applies it as it interprets it that time.
Where the document lives afterwards A saved project in your account. Reopen it in six months, edit it, export it again. In the conversation. Files stay downloadable within the conversation that produced them.
Long documents Pagination is deterministic code. A 60-page document breaks the same way a 5-page one does. Generated. Consistency across a long document depends on the model holding the whole thing in view.
Editing model Edit in place: click any text to change it, and insert headings, lists, images or buttons. Layout and pagination stay automatic. Re-prompt the model, or open the file in another application.
Contents page and pagination Built from your headings, with page numbers kept in sync. Whatever the generated document contains. Not maintained as the text changes.
Driving it from your assistant Yes. The MCP server lets Claude Code and Codex create and export documents in your account. Included on every plan. Not applicable: the assistant is the tool.
Pricing Free with unlimited exports (up to 5 content pages per ebook); Pro $16/month or $160/year. Included in a Claude plan.

You are not picking a side

The framing of this page is wrong, and deliberately so, because it is how the question gets asked. The useful setup is both: Claude writes, Unplain publishes.

Unplain runs an MCP server, so Claude Code and Codex can create and export documents in your Unplain account directly. You ask for the report, the model writes it, and it comes back paginated, branded and saved as a project you can edit. The writing stays with the thing that is good at writing, and the layout stays with something deterministic.

Which one fits your work

Choose Unplain if

  • You produce the same kind of document repeatedly and they need to match.
  • You revise. Editing text directly beats regenerating the file.
  • You want the document to still exist, and be editable, months later.
  • The output needs your colors, fonts and logo without you restating them.
  • The documents are long enough that pagination and a contents page matter.

Choose Claude on its own if

  • You need this document once and will not revisit it.
  • You need Word, Excel or PowerPoint rather than PDF.
  • The content matters and the presentation genuinely does not.
  • You want it in the next thirty seconds without connecting anything.
  • You are working somewhere Unplain does not reach, and the file just needs to exist.

Frequently asked questions

Can Claude make a PDF without Unplain?

Yes. Claude creates PDF, DOCX, XLSX and PPTX files in the web app, desktop and mobile, and it is genuinely good at it. For a document you need once and will not touch again, that is the faster answer and you should use it. Verified August 2026.

So what does Unplain add?

Three things, all of which only matter if you make documents more than once. You edit the text directly instead of asking the model to regenerate the file. Your brand is stored once and reapplied identically rather than described again each conversation. And the document stays as a project you can reopen, rather than living in the conversation that made it.

Why does editing directly matter if I can just ask Claude to change it?

Because a re-prompt is an approximation and an edit is not. Asking for a different opening line means the file is generated again: it costs tokens, it takes a round trip, and the model can quietly change spacing, a heading or a sentence you were happy with. Clicking the paragraph and typing changes exactly what you touched and nothing else.

Can I not just tell Claude my brand colors every time?

You can, and for one document it is fine. Across ten it drifts, because each conversation is an independent interpretation of the same description. A Brand Theme is a stored record rather than an instruction, so the tenth document matches the first because it is literally the same colors, fonts and logo, not a fresh reading of them.

Do I have to choose?

No, and the best setup uses both. Unplain's MCP server connects to Claude Code and Codex, so Claude writes the content and hands it to Unplain to lay out and export. You get the writing from the model and the layout from something deterministic, in one instruction.

When is Claude on its own the better choice?

When the document is a one-off, when you need DOCX, XLSX or PPTX rather than PDF, or when it does not need to match anything you made before. Unplain exports PDF only and its advantages are all about repetition, so a single document you will never revisit is not a case it wins.

Let Claude write it. Let Unplain publish it.

Start free. The MCP is included on every plan.