MojoPad eats almost anything. Every import becomes a page, named after the file, ready to link.
Notes scattered across two or three apps, and you want them together? Nearly every notes app can export what you've written as a folder — the notes in one place, the pictures they use beside them. Choose that folder with File ▸ Import…, or drop it onto a page, and the whole thing comes in.
[[Like this]] becomes
a real link to that page — MojoPad links by name, so once both notes are in, they find each
other.#tag written in a note is added to the page's own
tags, and stays in the text where you wrote it.You choose what comes in, and where each part of it goes. Most people's notes aren't one kind of writing — a journaling app often holds a diary in one journal, a notebook in another and a pile of half-thoughts in a third, and a folder of notes usually has folders inside it. So choosing an export doesn't import it: MojoPad opens on what it found, a line for each part with its own count, and you say where each one lands, rename it, or leave it behind entirely.
If your app offers a choice of export format, take the one that keeps attachments — some offer to put the pictures inside the notes themselves, which works here too. Export a handful of notes first and import those before committing to all of them; it takes a minute and tells you what you're getting.
Better said plainly here than discovered halfway through a library.
⌘F finds it in the reader (see Searching). Long
documents open instantly: pages draw as you reach them. You can mark passages as you
read — see Reading with a pen in your hand below. For ink-level markup, click
Open in PDF editor… above the page:
it opens in your default PDF app (Preview, Acrobat, …), and because the file lives
inside your document, your edits save straight back into it — press Reload
to see them..org), reStructuredText
(.rst), AsciiDoc (.adoc), LaTeX (.tex),
and log files. If it's really text, it opens as text.A page is named after the file, so the file’s name is the page’s name. For a
folder of papers that is not a detail: name each file after the work’s identifier —
PMID12345678.pdf, PMC1234567.pdf, or a DOI in its encoded form — and
every mention of that identifier anywhere in your wiki opens the copy you just brought in. The rule,
in full, with the reason a DOI cannot be a filename as it stands and working code for both
directions, is in Properties ▸ Naming a paper by its identifier.
Three kinds of reading, and MojoPad now handles all three without making you stop and take notes:
How. Open the PDF and select the words you want to mark. A small bar of colors floats just above the selection — amber, yellow, green, pink or blue, the same five the editor's text highlighter uses. Click one and the passage is marked; the color you chose is the one offered first next time, so marking a whole chapter in one color is one click per passage.
Changing your mind. Right-click a mark for its own menu: pick another color, or remove it. To clear every mark in the document at once, use the reader's toolbar. Marks live in the document, so they are there on every view of that PDF, they survive quitting, and they travel with the document if you move it.
Everything else the right-click menu did, it still does. Away from a mark you get the usual macOS menu — Copy, Look Up, Read Aloud — and it now works on whatever word is under the pointer, so you can look a word up without selecting it first.
Marking versus quoting. Dragging an excerpt into a note also leaves a permanent mark on the PDF, and that's the move when you want the words in your writing. Use a highlight when you want the passage to stay where it is and simply be findable — the point of the pen, rather than the scissors.
.webarchive in) offer a choice:
Clean & editable (the article, extracted) or Original snapshot (the
page exactly as saved, pixel for pixel — like VoodooPad kept them). The pictures come
with it. Anything the archive did not already contain is fetched once, as the page is
saved, and kept inside your document — so the page still has its pictures when the site
it came from has been reorganized or closed, and it reads the same on a train as it does
at your desk. If some of those pictures live on other sites, MojoPad names them and asks
once before fetching them; anything you decline stays as a link, and a page with links
left in it does contact those sites when you open it..html) import the same way — the same Clean
& editable / Original snapshot choice, with images and stylesheets from the
file's own folder pulled in so the snapshot still looks right. (Page scripts don't run —
snapshot freezes the static page, clean keeps the article.)Where a clip lands. By default, in whichever wiki you were last using — quick, and fine when you keep one. Keep several and it means what you clip while reading depends on which window happened to be in front, which is a filing accident waiting to happen. Set Clip to in Settings ▶ Web Clipper to a particular wiki and every clip goes there instead, whatever you have open; if that wiki is closed, MojoPad opens it, because a clip you cannot see is a clip you cannot trust. Underneath it, File clips under folder names the folder they arrive in — and that folder is created for you if it is not there yet.
Or choose while you clip. If you keep several wikis, a setting made once in the app is often the wrong shape — you know where a page belongs at the moment you read it, not in advance. Right-click the extension's button and choose Options, then check Ask which wiki, every time I clip: from then on the popup waits instead of clipping on sight, offering your wikis with the last one you chose already selected. (It arrives with an extension update, so if you do not see it yet, your copy has not been updated — the Clip to setting in the app works either way.) So the usual case is still one click, and moving a page somewhere else is one more. Leave it unchecked and clipping is exactly as it always was — the popup clips the moment it opens.
The list only appears when you have more than one wiki, since otherwise there is nothing to choose. And the browser is told what your wikis are called and nothing else — never where they live on disk, because a folder name can say more about your work than the page you are filing.
Getting the extension. It is free and part of MojoPad. Open Settings ▶ Web Clipper and click Get the Clip to MojoPad extension, which takes you straight to it; it runs in Chrome, Brave, Edge, Arc and Vivaldi. Safari cannot run it — on Safari the bookmarklet is the way, and it is on the same settings pane.
When a clip is not the page. A site can answer a fetch with a cookie dialog, a security check or a sign-in wall, and what lands looks exactly like a clip that worked — the title is right, the byline is right, and the article is nowhere in it. MojoPad measures how much of the page survived, and when very little did it writes a short note at the top saying so and what to do: clipping again as an Original snapshot, or clipping from the browser where you are signed in. Nothing is refused; the note only declines to let a clip pass for something it is not.
If your clipping is deliberately narrow — you have set things up to keep one part of a page and throw the rest away — that measurement is describing your own choice, and the note is worth turning off: Settings ▶ Clipping ▶ Warn on a page when the clip may not be the whole article.
Handing MojoPad a file from outside it. Right-click a file in the Finder and choose Services ▶ Add to MojoPad, print to Save PDF to MojoPad from any print dialog, or open a file with MojoPad: the file becomes a page, MojoPad comes to the front and opens it, and it lands where File imports under folder says — under Unfiled until you name a folder, so you can put it where it belongs.
Which wiki it goes to is Settings ▶ Files & Backups ▶ Importing files ▶ Add to. Left alone the file lands in the wiki you were last using. If you keep a wiki per project, choose Ask each time: MojoPad asks before anything is added, and the question offers the wikis you have open, the ones you have used lately, and Choose another wiki… for any other on this Mac. A wiki that is closed is opened for it; one with a password holds the file at its unlock screen and files it the moment you type the password. The question carries the folder as well, so one batch can go somewhere else without changing the setting, and Don’t Add adds nothing and leaves your file exactly where it was. Or name a wiki and every file goes there, whether or not it is open. This is a separate setting from Clip to on purpose: a paper you hand over is not a clip, and sending it wherever your clips go would put it somewhere you never chose.
A clip waits for the wiki, rather than being lost by it. If the wiki you clip to is closed, MojoPad opens it and holds the clip until it is genuinely ready — and if that wiki has a password, the clip waits on the unlock screen and files itself the moment you type it. If the wiki you named cannot be opened at all — renamed, moved, or on a drive that is not connected — MojoPad says so and offers to put the clip in a wiki that is open, rather than quietly filing it somewhere you did not choose. Nothing arrives silently and nothing is thrown away without telling you.
One case where dismissing is an answer. Out of the box MojoPad asks how to keep a whole page — clean and editable, or the page exactly as it looked. If you dismiss that question rather than answering it, the clip is discarded, and the status line says so, naming the page. That is deliberate, and it is safe for this one thing only: a web clip has another copy, because the page is still on the web and its address is still in your browser. It is also the alternative to what used to happen, which was the same question greeting you at every launch for a two weeks. Set a format in Settings ▶ Web Clipper and it stops asking altogether.
A clip that is not the article says so, at the top of the page it made. Some pages hand a clipper a consent dialog, a sign-in wall, or a shell that needed a browser to draw the words at all — and each of those files itself as a perfectly ordinary-looking page. You find out months later, when you open the one source you were sure you had. So MojoPad reads what it captured before filing it: if what came through is a cookie notice, a "turn on JavaScript" holding page, or so little text that it cannot be the piece, a short note at the top of the page says which of those it looks like. The clip is still saved and the note is ordinary text you can delete — it is there to be seen at the moment you file it, not to argue with you.
And it tells you in the browser, where you are. You clip from Chrome; MojoPad may be on another desktop, behind a full-screen window, or on a second display. Anything it needs from you there would be invisible — which reads exactly like a clip that failed. So when MojoPad is not the app in front and it has something to say about a clip, it sends a notification, which crosses desktops when a window cannot. Click it and MojoPad comes to you with the question waiting.
This is the one part of clipping that depends on something outside MojoPad: if you have turned MojoPad's notifications off in System Settings ▶ Notifications, there is nowhere for it to say anything, and a clip that needs an answer will sit waiting in a window you cannot see — the very thing this exists to prevent. Nothing is lost when that happens; bring MojoPad forward and the question is there. But if you clip a lot and never want the detour, setting a format (below) is the surer fix.
Most often that question is simply which format, because out of the box MojoPad asks each time. If you always want the same one, set it in Settings ▶ Web Clipper and it stops asking — clips then land without a detour, and the only notifications you get are the ones that matter.
Which setting reaches which. The pane says so beside each group, because the two tools do not take the same ones. Clip to and File clips under folder apply to both. Format for a whole page applies whenever a whole page is clipped, either way — clipping a selection never asks, because a selection is already just your text. Highlighting your pages as you browse needs the extension and does nothing without it.
Clipped articles arrive as ordinary rich-text pages: headings, images, and links intact, and any code or preformatted blocks render in tinted boxes that wrap long lines instead of running off the page.
Under any YouTube, Vimeo or TED video sits a small 📄 Transcript button — always there, no hunting for it. Click it. MojoPad fetches the video's captions and lays them out beneath the player as timestamped paragraphs — real, searchable text that becomes part of the page. Search finds talks by what was said in them; the transcript travels into your exports; and it renders wherever the video does, including on a canvas card.
The button does its work once and then goes away: a video that already has its transcript doesn't need it again.
Every timestamp is a link. Click one and the player jumps to that moment — no scrubbing. It's the fastest way to re-find the ten seconds you actually wanted.
Pull a quote onto a board — and jump back to hear it. Open the talk beside a
canvas (⌥-click it, or right-click its card ▸ Open beside the board), then
drag a transcript line onto the canvas. It lands as a sticky stamped
from 🎬 the talk · 12:34. Click that stamp later and the talk opens beside the
board, the video seeks to the exact moment those words were spoken, and a thread draws
itself from your note to the video. Your notes remember not just where an idea
came from, but when.
Transcripts are best-effort: a video with no captions, or a platform that changes its internals, may come back empty — an occasional maintenance patch keeps them flowing. Auto-generated (machine) captions are labeled as such.
⇧⌘7 in any application and a small capture
window appears; type a thought and it's appended to your Bucket page, timestamped. When
you capture from another app, MojoPad stays out of your way: the moment the panel
closes it hands focus straight back to wherever you were — your browser, your editor, the
call — so the capture never yanks you into MojoPad. Inside
MojoPad you can also open it from the 🪣 Bucket button in the toolbar or
File ▸ Quick Capture (Bucket). Click the 📌 pin in its title bar and the
panel floats over everything for drip-capture: each ↩ saves the thought and
clears the box for the next one, clicking elsewhere no longer dismisses it, and
esc closes it when you're done — perfect alongside a call or a video.⌥⌘V) — also on the page list's right-click menu.You have written a section that cites eleven papers, and you would like to know that it is true. Check What Your Sources Say… reads each cited sentence against the paper it cites and comes back with one of three answers.
Borne out — the source says this, and here are its words. Worth a look — the source's own words differ from what your sentence claims; they are quoted, and you decide. Not found — nothing supporting the sentence turned up.
That third answer is not the second one. Looking cannot prove a paper does not say something, only that this did not find it — perhaps in a section it did not read. Being told "your source disagrees" when the truth is "I could not find it" would be worse than being told nothing, so the two are kept apart. A source with no words to read — a scan that was never given a text layer — says that too, rather than being counted as a source that failed you.
Every quote is one the paper really contains. A model asked to copy a sentence will sometimes paraphrase instead, and a paraphrase inside quotation marks is a fabricated receipt on the one part of a claim a reader would never think to check. So a quote that is not literally in the source is discarded rather than shown with a caveat.
It will not rewrite your sentence. A button that softens a hedge on a machine's judgment would be the dangerous half of this idea — nobody re-examines a hedge they were told to add. It shows you the paper's words; the writing stays yours.
A locked source is not read, like everywhere else in MojoPad, and nothing is sent anywhere until you ask for the check.
You are writing a paragraph and you want three of your papers in it, each saying what it
actually found. Insert from Your Sources… does the whole of that without leaving the
sentence. Type /sources where you are writing; it is also under Edit and in the
command palette.
It opens with a search box that takes the whole query language: words, "a phrase",
OR, NOT, and the terms tag: in: type: name:
prop: and the date ones. prop: is new and is the one that matters for a
library: prop:Year=2025 finds papers with that year, prop:Authors=Kelly
finds an author, and prop:Title on its own finds every page that has a title —
useful mostly for finding the ones that do not.
Then there are three ways out, because there are three things a writer does with a shelf of papers. Link them puts every checked one in at once — as a list, or as A, B and C in the line you are writing — and needs no model at all; it is the quicker move, and the one for a see-also line or the sources behind a claim. Write them in takes an instruction instead — one sentence on its method and its main finding, a two-line annotation, what it says about older adults — and each source is read separately and written in as its own paragraph with a link back to it. Reference list is the third, and it is below.
An answer points at your own copies. When an answer names a paper you hold — by its DOI, its PubMed number, or the name of the page it lives on — the link opens that page rather than the publisher's website. Where you do not have a copy, the same link goes to the registry, which is the rule everywhere else in MojoPad: write the identifier once, and it finds your copy if there is one.
Reading one before you check it. Right-click any result for ⇄ Open Beside This Page, Open in Small Window or ＋ Add a Pane. The chooser steps out of the way while you read — a bar at the foot of the window says how many you have checked and what you are looking at, and Resume choosing brings the list back exactly as you left it. Nothing is lost by going to look.
What you type is what is asked. The instruction box passes your words to the model and adds nothing to them but the source itself. Ask for bullets and you get bullets; ask for a table, a heading, a numbered list, two sentences and no more — you get that. MojoPad adds only the link to each source afterwards, and the one rule it does keep is about honesty rather than shape: never invent a finding, a number or an author, and say so when the source does not answer.
One source at a time, on purpose — and all of them together when you ask for that. A model handed six papers at once blends them, and afterwards you cannot tell which sentence came from which — which is the opposite of what citing is for. So Write them in as starts at a paragraph about each source, in turn: each paper is asked about alone, and each paragraph belongs to one source and carries its link. Nothing lands on the page until the run ends — every source answered, or you stopped it — so you never see half a paragraph; a source that failed is left out and the run's log says so.
Some asks are about the set rather than about each paper: what the six of them agree about, a table of their methods, a list in a style this app does not ship. Those are one answer about all of them together. Every source is labeled in what the model is shown, so the answer can say which is which, and all of them are linked under it. Each source's share of the reading shrinks as you check more; where that would leave too little of each paper to count as reading it, the ask is refused and tells you roughly how many will fit — write the rest in as a second pass.
You can see it, and you can stop it. The moment Write them in starts, a run appears in Research runs & scripts and as a small card at the bottom right of the window, both saying which source is being read — reading 2 of 4. The row names the model that is answering, and so does the first line of the run's own log, so a small model standing in for the one you chose can never again go unmentioned. The card has a Stop, and so does the run when you open it from the list. Stopping ends the source being read and keeps the ones that had already answered. The run keeps a line for every source afterward: what it was read from, what was written, and why nothing was — a scan with no text, a model that answered with nothing, a source you stopped before it was read. The list shows how far a run has come each time you open it, and picking a run opens it: how long it has been going, which model answered, and every line it has printed, opened at the newest line. The card on screen keeps counting. Start a second job and its card takes the front; the first one's comes back, still counting, when the second ends. Check What Your Sources Say… and filling a property across a kind appear in the same list and stop the same way.
The end of a paper is its reference list, and reading a paper from both ends was handing the model the citations instead of the conclusion. Each source is now read from its beginning and from just before its references — the model is told the list was left out on purpose — and the run's log says when that happened and how far into the paper the list began. A paper with no such section is read exactly as before.
Reference list is the third way out, and it uses no model at all. Check your papers, choose Vancouver or APA 7, and each one is written from its own details — Authors, Year, Published in, Title, and Volume, Issue, Pages and Journal abbreviation where the page has them — read from its properties. A page that says too little but is named or labeled by a DOI, a PubMed number or a PubMed Central number is looked up through the same catalogs and the same kind of asking as ⌕ Look up this source: one question first, naming the catalogs and counting the requests — up to two for each page, one to the catalog and one more when its record names a DOI — and only the identifier leaves your Mac. What a catalog answers is used for the list and is not kept on the page; the lookup button is how to keep it. A field the record does not state is left out rather than guessed at; a journal is abbreviated only when the catalog supplied the abbreviation; and a page with too little to cite — no author and no title, or no year and no identifier — is left out of the list and named in the run's log instead of being printed broken. A page you have locked is left out too, and named the same way: its words are sealed, and a line in a numbered list that is only a filename is a gap wearing the shape of a reference. Vancouver entries are numbered in the order you checked them; APA entries are alphabetical and unnumbered, as that style is, and they hang their second line the way that style does. Each title is a link to its page.
An entry that has an identifier ends somewhere you can go. A Vancouver entry closes with the paper's DOI, or with the PubMed or PubMed Central number where there is no DOI. An APA entry closes with the address that identifier stands for — doi.org for a DOI, and the paper's own page at PubMed or PubMed Central for the other two — because a numbered entry carrying a title, a year in brackets and nothing else is a gap wearing the shape of a reference, and a paper known only by its PubMed number used to print as exactly that.
Two styles, and there is a reason there is no third. A named style like Chicago prints authors' full first names, and the catalogs disagree about whether they have them: one shelf of papers would come back with full names for some entries and initials for others, inside one list. It also wants titles capitalized word by word, which is a judgment about which words are names — and a list that is nine tenths right in a style you cannot check looks finished. Vancouver and APA 7 are the two this can do without guessing.
Two things it will not do, on purpose. It will not re-case an article title: which words in a title are names is a judgment, and a program that made it would make it silently and be wrong in both directions, so what you get is what the record says. There are two exceptions, both in APA and both of the kind that cannot be wrong: the first letter after a colon in a title, and the journal name, which that style prints in title case and which the catalogs hand over in sentence case — so The Cochrane database of systematic reviews prints as The Cochrane Database of Systematic Reviews. Only a lowercase word is ever raised, and only up: a word already written with a capital in it is left exactly as it was, so PLOS ONE, eLife and mBio come through untouched. A journal name that says it is in another language is left alone altogether — what says so is an accent, or a word like de, und or tijdschrift, so a name in another language that carries neither is capitalized like an English one. And it does not shape a book, a chapter or a conference paper; those print from the pieces they have. It is not the self-keeping References block, which lists what a page cites in no style at all — see A reference list that keeps itself.
Ask Write them in for a reference list in Vancouver format and it will not write prose about each source right away: a line under the box says that this asks for a reference list, the style is set for you, and Reference list is the button to press. Press Write them in again without changing what you typed and it writes the prose — so a question that happens to mention a paper's bibliography is pointed at the list once and then answered.
What it will not do is invent. The model is told to be faithful and to say when a source does not answer the question, rather than filling the gap. A source with no readable words — a scan with no text layer — is skipped and counted, so the number written in tells you if one was missed. And the model that answers is the one you chose: your own agent or the model you set up, when that is what the chooser says, and the run's log names it.
You want the address of a folder — to paste into a task manager, an email, or a page in another wiki. Right-click the folder in the sidebar and choose Copy Link to This Folder. What lands on the clipboard opens that folder in this wiki from anywhere on this Mac, with every folder above it opened too so it is genuinely on screen rather than merely expanded somewhere out of view.
That is a different thing from Link Selection to Folder…, which is in the editor's right-click menu and makes words in a page point at a folder. One puts a folder's address on the clipboard; the other writes a link into what you are writing.
Once a paper carries its numbers, they are listed under Aliases in the Info panel — and each one is a link. Click a DOI and doi.org opens it in your browser; a PubMed number opens PubMed, a PubMed Central number opens the article there. Right-click for Open in Browser, Copy Link or Copy Identifier, which is the quickest way to paste a paper's DOI into a message.
Nothing is fetched by MojoPad when you do this — the address goes to your browser, which is the same thing that happens when you click an identifier written in your own prose.
You have a shelf of offprints and scans, and half of them carry no DOI anywhere in the document. That is normal — a manuscript accepted for press, a chapter, anything published before identifiers were universal — and it used to mean the paper could never be looked up at all.
Now the title is the key. ⌕ Look up this source appears on those pages as well, and selecting a batch offers to search by title for the ones with no identifier — telling you how many before it asks, because a title is words out of your library and an identifier is not. The title goes out, and the first author's surname where the page keeps one — which is there to tell two works of the same name apart rather than to make the match. Nothing else does.
It matches exactly or not at all. A catalog will rank five plausible near-misses above nothing, and a wrong identifier here does not stay cosmetic: it becomes an alias, and starts linking your writing to somebody else's paper. So a work is taken only when its title matches yours word for word — capitals, punctuation and typesetting aside — and where two works answer to the same title with nothing to tell them apart, neither is taken. At the end of a batch the ones identified this way are counted on their own line, because they are the ones worth a glance.
And the paper it finds brings its other numbers with it. A title match ends in a DOI, and a DOI is then asked of PubMed for that paper's PubMed and PubMed Central numbers. All three land as properties — DOI, PMID, PMCID — and each one also becomes an alias, so writing any of them in a note links to the page and a citation written any of those ways opens your copy.
A paper whose file name is a DOI with underscores where the slashes should be —
10.1093_ptj_pzaa150.pdf, which is how a fetching program writes one when it cannot put
a slash in a filename — cannot be read as an identifier. Real DOIs contain underscores of
their own, so nothing afterwards could tell which was which, and a rule that swapped them back
would file some papers under the wrong work in silence. So MojoPad asks instead: pressing
⌕ Look up this source checks up to four spellings of that name with the registry
— an identifier-shaped guess, never your words — and takes one only if it is the
only spelling the registry knows — a record without a title still counts as a
spelling it knows, because the name is ambiguous either way. Where the page already carries a
Title, or a name that reads as one, the record's title has to match that as well —
which for a paper freshly imported under a name like this one is usually not yet the case, so on
those pages the registry knowing exactly one spelling is the whole of the check. The button says so before you press it, and the page's name
is left exactly as it is. Next time, name the file with
the one your assistant's read_identifier gives you and none of this is needed.
Copy Link to Selected Words works in a stored PDF as well as on a page. Select the sentence, right-click, and take the link — the same command, by the same name, in both places.
A page can keep the mark by wrapping the words in the text itself. A PDF cannot be written into, so the words are highlighted where they sit and the link points at that highlight — which is also how citations into a paper have always worked here, so the mark is a thing you can see and remove afterwards. Following the link opens the paper beside what you are reading, scrolled to the sentence with the highlight showing.
The link works anywhere on this Mac. Unlike the page version it does not offer a web address: a highlight is MojoPad's own idea of a place, and a shared copy of the wiki has no way to jump to one — so rather than hand out an address that quietly lands at the top of the paper, it does not offer the choice.
A project note wants to point at the folder its papers live in, not only at one of them. Select the words, then Link Selection to Folder… — in the right-click menu and under Edit ▸ Link, beside the same command for a page. Pick any folder in Collections and the words point at it.
Clicking such a link opens that folder in the page list — and every folder above it, so it is actually on screen rather than expanded inside something still closed — and shows what is in it as a gallery, the same view the folder's own ▦ button gives. The link is stored by the folder's identity rather than its name, so renaming the folder does not break it, and two folders sharing a name are never confused for one another.
You have downloaded four hundred papers and every one is named after a fragment of its address. A paper fetched by its identifier arrives as pnw107.pdf — the last piece of 10.1093/pm/pnw107, because a web address had nothing else to give. Nothing truncated it; the title was never in the filename. So the one thing you would search for — what the paper is about — is written nowhere MojoPad can see, and the folder of them reads as a list of serial numbers.
Settings ▸ Files & Backups ▸ Importing files answers the first half. Add to names the wiki a file goes to when you hand one to MojoPad from outside — the Services menu, Save PDF to MojoPad, or Open With — so a bulk import does not depend on which window happened to be in front. File imports under folder names the folder they go into, made for you if it is not there. Name a path and it nests — PDF Library/Endoscopy files two deep and creates whichever part is missing, so a library can grow subfolders without any of them being made by hand first. Leave it blank and they stay at the top level, which is what MojoPad has always done. On whichever wiki is in front, or with a wiki named, nothing is asked at the time: the point of a setting here is that a hundred files can go in at once without a hundred dialogs — change the setting between batches and each batch lands where you last set it. On Ask each time you are asked once for the whole batch, and the answer names the wiki and the folder both. The folder box in that question starts on this setting until you have answered it once; after that it starts on the folder you chose last time, so twenty papers into one project is one answer and nineteen agreements. Clearing the box is an answer too — it means the top level, and the next question starts blank rather than back on the setting.
Where a batch went, and a second chance to move it. Files handed in from outside announce
themselves once when the batch is done — Filed 12 files into “PDF Library” — with a
Move… beside it. Choose it and you can send that whole batch somewhere else; typing a path
that does not exist yet makes it, so the batch that belongs in a new subfolder gets one without
you creating it first. The move can be undone. The notice goes on its own after twenty seconds; the
✕ closes it at once, and so does esc — you no longer have to start a move
you did not want in order to be rid of it.
A batch is one question, never one per file. The Finder service hands MojoPad one file at a time, so a question at each arrival would be a question per paper. Files are collected as they arrive and asked about once, naming how many — and on the other two settings nothing is asked at all, so a bulk import runs from start to finish uninterrupted.
The second half is the name. A PDF usually knows its own title — publishers write it into
the file — and MojoPad now reads it as the words are read, and puts it in two places: a Title
property, which is where you read it and what a citation and a database view use; and an
alias, which is what makes it findable. From then on the page lists under its title rather
than its filename, typing [[ finds it by title, and [[The Paper's Full Title]]
links to it. The filename still works too — nothing you could type before stops working.
And the identifier a paper prints inside itself. A journal article states its own DOI on
its first page. That is picked up as an alias as well, so [[10.1093/pm/pnw107]] links to
the paper, and Look up details can find the work — which it could not do before, because it
reads a page's name, aliases and properties, and all three said pnw107.pdf.
What it will not do is guess. Plenty of PDFs carry junk where the title should be: a word-processor filename, untitled, or a typesetting job number. A wrong title here would not be a cosmetic slip — it becomes an alias, so it starts matching your prose and linking your writing to the wrong paper, and it becomes what the page is called. So anything that does not read like a title is refused and the page keeps its filename. No title is a better answer than a confident wrong one, and for a paper with a DOI you can ask for the details to be looked up and get the real one from the registry.
Nothing already there is overwritten. A title you set is yours, a page you renamed keeps your name, and an alias that already names another page is left alone rather than made to mean two things.
File ▸ Merge Document… imports every page of another MojoPad document into this one — names are de-duplicated, links are rewired, and each page keeps its kind and the fields that go with it, the definitions they are read through included. File ▸ Split Document… goes the other way: pick pages, save them as a new document, optionally remove them from this one.