Skip to content
Multilo Docs

Common workflows

Step-by-step recipes for everyday research writing: exploring sources, drafting, editing, citing, reviewing, and working with Word files, data, and media.

This page collects short recipes for everyday research and writing work in Multilo. For higher-level advice on prompting and context, see Getting better results.

This page covers:

Prompt recipes

These are prompt patterns for everyday tasks. Type them in chat with your project open, and change the wording to fit your own work. Each step builds on the one before it.

Get an overview of your sources

Suppose you have just started a project, or picked up someone else’s, and need to understand what is in the library quickly.

  1. Open the project

    Open the project folder in Multilo, or create one and drag your papers in. See your first project.
  2. Ask for a high-level overview

    “Give me an overview of the papers in my library.”
  3. Dive deeper into what matters

    • “What methods do these studies use?”
    • “Where do the authors disagree?”
    • “Summarize the main finding of each paper in one sentence.”

Tips

  • Start with broad questions, then narrow down to one theme.
  • Ask for a short glossary of the key terms in the field.
  • Add missing papers with Sources Hub first — the overview is only as complete as your library.

Find a passage

Suppose you remember an idea from your reading but not where you read it.

  1. Search by meaning

    Open Super Search, turn on by-meaning search, and describe the idea in your own words.
  2. Or ask chat to find it

    “Find where my sources discuss response bias in online surveys.”
  3. Jump to the exact sentence

    Click a result to open it in the document view with the match highlighted, then ask a follow-up about it.

Tips

  • Use the terms your field uses, not everyday words.
  • Search looks inside PDFs, Word files, and PowerPoint decks, with page and slide badges.
  • Turn by-meaning off when you know the exact phrase you are looking for.

Draft a section

Suppose you need to write one section of a paper from your notes and sources.

  1. Ask for an outline first

    “Outline a Related Work section from the sources in my library. Don’t write it yet.”
  2. Draft from the outline

    “Draft the Related Work section from that outline, about 600 words, citing sources from my library.”
  3. Refine one thing at a time

    “Make the second paragraph more concrete and add a counter-argument at the end.”

Tips

  • Give a target length and name the section you mean.
  • For a whole paper rather than one section, use the Full Draft Writer.
  • Set up voice.md once so every draft sounds like you.

Edit and polish your writing

Suppose you have a rough draft that needs to be tighter and clearer.

  1. Ask for feedback before edits

    “Which three paragraphs in this section are weakest, and why?”
  2. Make a targeted edit

    Select the introduction first, so the edit lands in the right place, then ask: “Tighten this introduction to about 150 words. Keep the thesis in the first sentence and don’t add new claims.”
  3. Run a focused agent

    Open the Agents Library and run a specialist on the section — for conciseness, plain-language rewriting, or humanizing prose that sounds machine-written.
  4. Keep polishing as you write

    Switch inline suggestions to Automatic and accept or dismiss each proposed edit as it appears.

Tips

  • Edit in small steps, so each change is easy to check.
  • Pick a light MODE for a quick pass and a deeper one for a thorough rewrite.
  • Every change can be undone from history while you decide.

Fix and check citations

Suppose your citations are in mixed styles, or you are not sure every one supports its claim.

  1. Put everything in one style

    “Reformat every in-text citation in this document to APA, and flag any that don’t match a source in my library.”
  2. Check that each source supports its claim

    Run Claim Check. It fetches each cited source and tells you whether it supports the claim, is weak or unrelated, or has a stronger alternative in your library.
  3. Swap in stronger sources

    Accept a suggested replacement and the in-text citation and the reference update together.
  4. Make the reference list match

    “Make the reference list match exactly what I cite in the text.”

Tips


Review a draft before you share it

Suppose you want a critical read before the draft goes to your supervisor, co-authors, or a journal.

  1. Start a fresh chat

    A new chat gives the AI a clean slate, so it reads the draft without the conversation that produced it.
  2. Check the method

    “Review the methodology section for sample size, controls, and how the statistics are reported.”
  3. Look for gaps in the argument

    • “What would a skeptical reviewer say about this argument?”
    • “List every claim in the Discussion that doesn’t have a citation.”
  4. Fix, then verify

    Make the changes you agree with, then run Claim Check as the last pass.

Tips

  • Use a flagship model for critique — see choosing a model.
  • The Agents Library has a methodology review agent for exactly this job.
  • Ask for the problems first, and decide on the fixes yourself.

Work with Word documents

Suppose you have a .docx to improve, or you have to hand your work in as a Word file.

  1. Open the Word file

    Drag the .docx into your project. Multilo reads its headings, lists, tables, footnotes, and citations and opens it in the editor.
  2. Improve it with chat and agents

    “Tighten section 3, improve the flow between paragraphs, and reformat the citations to Harvard.”
  3. Save back to Word

    Save as .docx. The reader gets an ordinary Word document with your edits applied and the formatting intact.

Tips

  • Add the sources the document cites to your library if you will check or extend its citations.
  • For the full walkthrough, see Edit a Word document with AI.

Write from your data and results

Suppose you have a folder of results — exported data, tables, figures — and need to write about what it shows.

  1. Put the results in the project

    Drop the folder into your project. Once it is there, agents can read it. See files & folders.
  2. Ask what the results show

    “Summarize what the files in the results folder show.”
  3. Turn data into a figure

    “Make a chart of the scores in results/scores.csv.” The agent writes and runs code in a sandbox to make it. See tools & Agent Compute.
  4. Write it up

    “Write the Results section from these files, with the key numbers in the text.”

Tips

  • Keep results in a clearly named folder so agents find them.
  • Ask the agent to compare two sets of results and describe the difference.
  • Check every number in the text against the source file.

Work with images, audio, and video

Suppose your research includes figures, scanned pages, recorded interviews, or lecture clips.

  1. Attach the file in chat

    Attach an image, an audio recording, or a video to your message. Images also open in the built-in Image Viewer.
  2. Ask about an image

    • “What does this chart show?”
    • “Describe the trend in this figure in one paragraph for my Results section.”
  3. Work from a recording

    “Transcribe this interview and pull out the main themes, with a short quote for each.”

Tips

  • Use an image when describing it in words would be slow or unclear.
  • Transcription runs on your computer. See media & images.

Point at files and folders

Type @ in chat to point at a specific file, folder, or source, so the answer is grounded in exactly the material you mean.

  1. Point at one file

    “Summarize the argument in @smith-2021.pdf.”
  2. Point at a folder

    “What’s in @results, and what does it tell us?”
  3. Compare several

    “Compare the methods in @smith-2021.pdf and @lee-2023.pdf.”

Tips

  • You can mention several files in one message.
  • Can’t remember the name? Describe the file instead and the agent finds it.
  • One well-chosen file beats ten loosely related ones.

Pick up where you left off

When a task spans several sittings, you don’t have to explain it all again. Your work lives in the project, not in a chat: open the project and your document, library, and .multilo/ rules are all where you left them.

To carry a fact forward, ask an agent to remember it:

Remember that this paper uses British spelling and APA style.

It is saved as a small note in agent memory, and agents recall it in later chats. Rules you want applied every time belong in AGENTS.md. See agent memory.

Keep writing while an agent works

You don’t have to wait for an agent to finish. Several sessions — you in one window, an agent in another — can work on the same document at once, and their edits combine instead of overwriting each other. Write the introduction yourself while an agent drafts a later section.

If a change turns out to be wrong, restore an earlier version from history: the last 40 versions of each document are kept for 14 days.

Plan before you draft

For big changes, ask for the plan first and review it before anything in your document changes:

Propose how you would restructure chapter 2 so the argument builds step by step. Don't change anything yet.

Read the plan, adjust it, then ask the agent to go ahead. The Full Draft Writer works this way on its own: you see the outline land first, then the sections fill in.

Hand a big job to an agent

For a whole-document job, describe the outcome in chat and let it pass the work to the right agent:

Write a 3,000-word literature review on remote-work productivity from the sources in my library, organized by theme.

Chat routes a long write like this to the Full Draft Writer and asks you to confirm the project before it starts. You can also type /draft to start a full draft directly. See Run a literature review.

Turn a repeated task into an agent

If you find yourself asking for the same thing again and again, save it as your own agent with /createagent:

/createagent a "Methods reviewer" that checks a paper's methodology section for sample-size, controls, and stats reporting

The new agent appears in the Agents Library with a YOURS chip, and you can run it on any document in one click. See custom agents.

Next steps