Tangle
A plot checker for novels. Each card is a scene, and scenes sit in chapters. Tangle tells you when a scene depends on something the reader has not been told yet.
Write three things on each scene: what it needs (what the reader must already know), what it reveals, and who is in it. That is enough for every check.
Break: a scene needs something nothing earlier reveals. On the Example tab, tap chapter 9 and drag it one place left to see one, or cut chapter 10 and watch the ending break. Warning: three scenes in a row where no main character’s story moves. Note: worth a look, not broken, such as two scenes in a row where nobody’s story moves, or a reveal that nothing later uses.
Your characters. Tap one to read their arc in story order. Open a scene to see who is in the room, and what they were there for.
Not sure yet? Chapters 4 to 6 of the Example are not in order, and Tangle checks every order. + other version on a scene plans a second way forward, so two endings can sit side by side. A draft (in More) holds a what-if without touching your plan.
“Start your book” on the sample site opens a board already set up in chapters, with ready outlines for a mystery, a romance, a thriller or three acts. Any board can switch with More → For: Story. The rest of this page is for game masters running D&D. The checks are the same.
A prep checker for people who already run games. It draws your campaign as a graph and tells you where it breaks.
You already know how to write a session. What nobody can do reliably is hold every branch at once and notice that one of them is broken. That is the job this does.
You describe scenes, what each one needs, and what it hands over. It walks every route your table could actually take and reports the ones that do not work. It reads dependencies. The writing, and the taste, are yours.
Four tabs. Your story is your own board — two scenes to start, saved in this browser. D&D example is a finished eleven-day heist and Story example is a short mystery novel; pull either apart without touching yours — every board is a separate save. How it works is this page. The Game / Story switch above the board changes only the words: the party or your characters, sessions or chapters.
One HTML file, no account, no install, no network request of any kind. It runs at the table with the wifi off. Save to file writes the whole campaign to a file; Open file takes one back — that is your backup, and how you hand a game to a co-DM. Both are under More, above the board.
A linear campaign needs none of this. But the moment the party can go left, your prep stops being a document and becomes a graph — and every path through it has to stand on its own.
The failures are silent, because each piece is individually fine. You put a letter in the east room. Four sessions later an NPC asks them about the letter. That works — unless they went through the west room, which they did, because you gave them the choice and meant it. Now you are live at the table and your own notes are asking a question the party cannot answer.
Nothing in your prep is wrong. The mistake exists only in the combination, and a campaign with a dozen real choices has more combinations than anyone can hold in their head at 7pm on a Thursday.
It is spellcheck — for a document that can be read in eighteen orders, where the typo only appears in nine of them.
Which means the thing it is actually for is the improvising table: the more your players go sideways, the more paths exist, and the more of your prep has never been checked by anyone.
The wall. On some route, a scene gates on a key that route never handed over. The single most common thing wrong with branching prep.
Your villain is hostile in act three and no scene anywhere turns them hostile. You know why; the prep does not say. Obvious in hindsight, invisible while writing.
The party can spare the prisoner or let him be killed. Two scenes later he is on stage, talking. That works on one of the two routes, and the board says which.
Somebody has been chasing a personal thread for four months and this particular ending never gives them the scene that closes it.
A scene billed as wide open where every possible action leads to exactly the same next scene. A railroad you did not mean to build.
Three approaches that look independent and all quietly route through the same NPC. Kill them, or lose their trust, and all three shut at once.
Everyone had fun and nobody got anywhere. Better to know on Wednesday than on the drive home.
Measured from the sessions you actually played, not from what you hoped. This is the number that decides whether the finale lands before summer.
Chekhov’s gun. Something handed over that nothing ever asks for again. Might be a slow burn, might be a dropped thread — it reports, it does not judge.
A corridor. Not wrong; most campaigns have some. But four sessions running where nobody chooses anything is very hard to feel while writing them one at a time.
The other two are refusals. On the page it reads as a choice. At the table it is a door and two walls, and they will notice.
The thing that stops it cannot exist before session seven and it fires on session six. Invisible on paper, because you wrote the deadline and its counter weeks apart in different files.
You decided what the act was for and then prepped scenes that do not do it. Same sentence as the wall, one size up.
The row of buttons above the board — Scene · Session · Arc · Act — redraws it in bigger units. One card per evening, or per chapter. The campaign is the thing all four sit inside, not a fifth setting: a board showing a different campaign is a different document, not a wider zoom.
It is not a summary. It is the same question asked about a bigger box: what does this chunk need that nothing inside it provides, and what does it hand on to everything after? That is exactly what a single scene is asked, so the answer reads the same way at every size — and a chapter that quietly depends on something you never established shows up in the same words as a scene that does.
It is also how you plan from the top. Say “by the end of act one they have the ledger” before you have written anything, and the board will keep telling you that act one owes you a ledger until some scene inside it hands one over. Or ignore all of it, start with scene one, and see where it goes — a chapter that promises nothing is never in debt.
At Arc and Act there is also a card called not in an act yet. That is not a complaint. Near the start of a plan, the pile of things you have not filed is the most useful thing on the screen.
Middle row: the story, left to right. Where the boxes fan apart, the group had a choice, and the branches are the different ways it can go. Some rejoin. Some run to their own ending — the ones with a star.
Top: one coloured string per player, dipping down to touch every scene where their story moves. Each string comes in from off the left and ends in a loose knot, because nobody starts existing at their first scene. A string that doubles back on itself is a beat happening before the thing it depends on.
Bottom: one line per agent, coloured by how they feel about the group. Where the colour changes, something changed them, and the board says what.
Each string sits in a faint tinted channel that runs exactly as far as that character does — where the channel stops is the information. Scroll past the end of somebody and their lane closes and the ones below slide up, so the board only ever carries the cast still in play, sorted with the busiest nearest the story.
Above the scenes: bands. Nights at the bottom, then arcs, then acts, then any deadlines. A deadline’s band spans exactly the room the group has — from the first moment the thing that stops it could exist, to the night it lands — with a dashed wall down through the story where it falls. If the band is hatched and reads backwards, there is no room; they cannot get there.
Underneath it all: the list of everything currently wrong. Click any of it to jump to the scene.
Click a scene and two things happen. The panel opens on the right, and the lanes collapse to just the characters that scene involves — the ones who appear in it, whose beat lands there, whom it changes, or who supply the way in or out. Somebody tied to it without being on stage gets a dashed tether instead of a knot, because “involved here” and “in the room” are different facts.
Everything above describes reading a finished board. Yours starts as two scenes and a title, and you build it by clicking whatever you want to change. No edit mode, no save button — the board is the document and it writes itself to this browser as you go.
The order below is roughly how prep actually happens, but nothing depends on following it. A half-finished board is the normal state of a board, and the findings are written to be useful while it is still wrong. You can also port an existing campaign in a scene at a time; the checks get sharper as you fill in Needs and Hands over, and cost nothing until then.
The title at the top of your own board is a text field. Click it and type. That name is what the exported model and the copied brief are filed under.
Click a card and the panel opens on the right: a title, a paragraph, and the handful of fields that make it checkable. + after puts the next scene down the line from this one.
A branch is a fork the tool can reason about — it makes the options nameable, so the picker can walk one way and not the other. This is the difference between a story that has choices and a board that knows it does.
Needs is what it cannot start without; Hands over is what the party carries out; Makes true is a fact nobody carries. These three fields are where almost every wall comes from, and filling them in is the whole price of admission for catching one.
A scene with something in its Ending field is where a route can stop. Until at least one exists, every thread just trails off — which the board will tell you about, because a scene with nowhere to go and no ending is a loose end rather than a finish.
+ new player and + new agent in the Cast tab under the board. Then, on any scene, + new beat lands a piece of somebody’s personal story there, and + put an agent on stage says who is in the room and how they feel about the group.
+ change records that somebody’s stance moves — on a scene, or on a route out of one, where it can be gated so it only fires if they went that way. An agent who is hostile later and never turned hostile is the finding this prevents.
+ act and + clock are in More, above the board. An act can promise something before a single scene inside it exists, and the board carries the promise as a debt until one delivers. A clock lands on a night whether or not anyone is ready.
Nothing has to be finished. A route you have not worked out yet is drawn dotted and stays on the to-do list; a fork with one arm written is a fork with one arm written. What you do get, from the first two scenes onward, is that the list under the board recomputes on every change — so a plan breaks in front of you while you are writing it, which is the only time it is cheap to fix.
It will not write anything, generate anything, or suggest anything. It has no opinion on whether your villain is interesting, whether the ending is earned, or whether the table is having a good time — those are your job and it cannot help. It reads structure: what depends on what, and nothing else.
Every judgement it makes is one you could make yourself with a very large sheet of paper and a free afternoon. It is just faster, and it never gets bored halfway down the page.
Fifty years ago the Reliquary District burned. The record of who was paid to let it happen was sealed for a century. The seal breaks in eleven days.
Three hundred people died in the Ashfall, and the inquest called it an accident. It wasn't. The Cinder Ledger — the register of every bribe taken to keep the fire-wardens away that night — was sealed by court order in the Ashvault beneath the burned district, in the custody of the guild that keeps such things. In eleven days the century lapses. Everyone with a reason to read it first, or burn it first, is already in Corvane.
Edges carry confidence and triggers, never content.
Three ways down is three routes, not three scenes. The front door, Oleth's stair and the Chain's tunnel all land in the same room and you'd prep it once — what differs is who the choice costs you. That's the rule: if you'd prep it separately it's a scene; if you'd improvise it, it's a route. On the board they're three arcs between the same two nodes, each labelled with what it turns.
Dotted routes are the ones where you don't yet know how the party gets from here to there — which makes them a prep queue rather than a defect. There are three: S3→cluster, S7→gate, and S13→S14, the last being the biggest unsolved jump in the adventure.
The fixture is built to be broken in specific ways.
Aurelia's string folds back on itself: she testifies about a signet she hasn't recovered. The literal tangle, visible without reading a word.
Add "the proxy recognises the signet" and his track turns red where it should. Three mismatch markers clear at once.
Two findings vanish, the dashed string segments go solid, and the shape of the board narrows. Story mode and adventure mode, same data.
One of the three descent arcs disappears and the guild becomes the only lawful way in. Removing an ally should be visible as the board getting thinner.
Narrow, then wide, then narrow. Nobody typed that — it falls out of the route graph, which is the argument for computing Agency instead of authoring it.