Use AI assistant to format paper documents in MCP

Key points of this article

Organize the school's formatting requirements into a list and give it to the AI assistant as a basis for layout. It is much more stable than just saying "arrange the highlights"
After accessing through MCP, the AI assistant can create and modify Word or PDF files instead of just giving text suggestions
Let the assistant check the structure first, and then modify the format in rounds. Only one type of content is changed in each round, making it easier to locate problems when they arise.
Header page numbers, table of contents, figure numbers, and references must be opened and checked one by one. You cannot just read the assistant’s reply.
For content involving machine translation or OCR, you must read it through yourself before submitting it, focusing on checking formulas and numbers.

If you want to submit a graduation thesis or journal manuscript, what takes time is often not writing, but adjusting the format according to the requirements of the school or journal: page margins, title levels, line spacing, table captions, header page numbers, and reference styles. Changing one place often affects several places. Most online tutorials teach you to set styles manually, but if you already have a draft and just want to hand over the typesetting work, a more practical question is: Can the AI assistant directly modify the file?

The so-called paper formatting skill can be understood as organizing a set of formatting requirements into fixed instructions, so that the AI assistant can execute according to the same rules every time. This article is written for graduate students and research group members who already have a draft of the paper (Word or PDF) and want to use AI assistants to complete the typesetting. After reading, you will know how to access OmniDoc's MCP service, how to write the format requirements into a basis that the assistant can execute, and which areas must be checked by yourself.

How to use the paper formatting skill? One sentence conclusion

First connect to the MCP service so that the assistant can directly create and modify files; then write the school format requirements into a list, ask the assistant to execute it in rounds, and open the file to check after each round. The table below provides specific methods based on your situation.

your situation	Recommended practices	reason
The paper is a draft in Word and must be formatted according to the school template.	Organize the template requirements into a list, let the assistant check the title level and paragraph style first, and then modify them collectively.	Word’s style structure is clear and changes can be traced
Only PDF version, you need to change some text or add signatures	Use the PDF editor to change paragraph text, sign or fill in forms, and let the assistant process the page.	Directly changing the PDF layout has a higher risk, but local modifications are more controllable.
Scanned PDF or foreign document translation	First do OCR or translation, then typesetting, and finally check it paragraph by paragraph.	Both recognition and machine translation may make mistakes, so check them before and after typesetting.
Both Word and PDF are required before submission	First finalize the Word version, then export it to PDF, and finally use the comparison function to compare the old and new versions.	Avoid inconsistencies in the contents of two files

Scenario: A 30-page paper is to be submitted in school format

Imagine that you have a draft of a master's thesis of about 30 pages. The content is basically completed, but the school has more than a dozen format requirements: A4 paper, text size 4 in Song style, line spacing set according to school regulations, first- and second-level title numbers are unified, table captions have a fixed format, references are arranged according to school regulations, and header page numbers are counted from the main text. These requirements are dispersed in school notices, template pages, and verbal reminders from instructors.

What you will deliver is a submittable Word document, plus a PDF. There are usually three places where things get stuck: the title level is confused due to copying and pasting, the table of contents page numbers do not match after changing the font, and the chart numbers are not updated simultaneously after adjusting the chapters. If you throw these to the AI assistant at once, it may correct one part and miss another, so a sequential approach is needed.

Complete process

Organize the format requirements: Write a list of the requirements in the school notice and template, including paper, page margins, font size, line spacing, title level, table captions, header page numbers, reference format, and attach a screenshot or file of the template page.
Connect to OmniDoc MCP: In an AI assistant that supports MCP (such as Claude, Cursor, etc.), follow the MCP service instructions provided by OmniDoc to complete the addition, and then confirm that the assistant can call file-related operations.
Ask the assistant to check the structure first: After reading the draft, ask it to list the order of chapters, titles at each level, the number of tables, and the number of references. You can check the content to confirm that it is correct before continuing.
Unify title and text styles: Let the assistant process the title level and body paragraphs first, change only this type of content, and save the copy before proceeding to the next step.
Processing page settings and line spacing: Paper, page margins, and line spacing are completed in a separate round. After completion, a few pages are randomly checked to confirm that they have not affected the previously unified titles.
Check the figures and references: Ask the assistant to list all the figure captions and the citation numbers in the text, and compare them one by one. If there is any inconsistency, point out the page number and ask him to change only that part.
Generate PDF and compare: Generate PDF from the finalized Word. If there is an old version, use the compare function of the PDF editor to find the differences, and then check the header page number and signature page.
Submit after manual reading: Read the entire text, focusing on formulas, numbers, units and citation years, and confirm that it is correct before submitting.

Where things can easily go wrong

Assistant reply "completed" does not mean the format is correct. The way to find out is to open the file, randomly select 5 pages and look at the list one by one, especially the table of contents, page numbers and section positions. The solution is to tell the assistant the specific description of the problem page and ask that only that part be changed and not re-typeset the entire text.
Too many requests are made at once. Font changes will affect paging, and paging will cause the table of contents page numbers to be misaligned. The processing method is to execute in rounds, and save a version in each round. When problems occur, you can roll back to the previous version.
Title hierarchy confusion. The common reason is that other formats are brought in when copying and pasting. The way to find out is to check whether each chapter title uses a unified title style, rather than just looking at whether the font size is consistent. The solution is to ask the assistant to apply the title style uniformly, and then check the number.
The content of the translation or scanned copy is incorrect. OCR and machine translation can misread symbols, formulas, and proper nouns. The discovery method is to compare the formulas and numbers one by one with the original text, and the processing method is to manually correct them before typesetting them, rather than changing them after typesetting them.
Upload unpublished data. Before handing over any online service, read its privacy statement; sensitive data can be processed on a local copy first.
School Format Update. The format requirements are subject to the school’s new notice. Confirm the version before starting work to avoid formatting according to the old template.

🎯 Which method is suitable for you?

If you already have a draft in Word and just want to unify the format

Connect to the MCP first, and let the assistant modify it in turns in the order of title, text, page settings, and charts. Save a copy in each round, which is safe.

If you only have PDF, you need to change some text or add signatures

Prioritize using the PDF editor to directly modify paragraph text, signatures or filling out forms, and only leave it to the assistant when you need to adjust the page structure.

If you want to reuse formatting requirements for multiple papers in the same institution

Save the format list and give it to the AI assistant as a fixed basis each time, and check it individually before submitting each paper.

If you are not confident about AI assistant operating files

Work on the copy first, keep the original, check and save the version in each round, and then replace the official document after confirming the effect.

❓ Frequently Asked Questions (FAQ)

1. What exactly is paper formatting skill?

It is a set of practices that organizes the formatting requirements of schools or journals into fixed instructions and allows the AI assistant to execute them according to the same rules every time. The key is to write the list in a specific way, such as "The main text should be in Song style and the line spacing is in accordance with the school regulations", which is much more reliable than "arrange it more nicely".

2. Can I use this method if I am not connected to MCP?

Can. You can give the formatting list to any AI assistant and let it give you suggestions for each step, and then you can set it item by item in Word. The difference is that after connecting to MCP, the assistant can directly modify the file, eliminating the need to copy and paste back and forth.

3. Can documents typewritten by AI be sent directly to the school?

It is not recommended to submit it directly. According to the verification steps of this article, check the table of contents, page numbers, figure numbers and references page by page, and manually read the full text and confirm that it is correct before submitting.

4. Can scanned papers be formatted using this method?

Yes, but you need to do OCR first and then typeset. There may be errors in the recognized text, especially formulas and numbers that need to be checked against the original text.

📝 Summary

The difficulty of paper formatting is that there are many formatting requirements and changes are interrelated. Giving the format list to the AI assistant connected to MCP, revising it in rounds and checking it item by item is a more secure way of formatting the paper. The value of paper formatting skills is to make each round of changes well-founded and reversible. In the end, manual reading is still required to ensure the quality of the manuscript before submission.

Open OmniDoc and connect the documentation tool to your AI assistant