KNOWLEDGE GRAPH MAKER
The connections are
the knowledge.
Notes and sources as nodes, labeled relationships between them, and the holes in your understanding visible as empty space rather than as something you have not noticed yet.
Free plan coming late 2026
Works in your browser
No account needed to view
Used by creative professionals at:
Artlist
Pixar
Nike
Red Bull
The North Face
Porsche
Start from a ready-made template
Pick a board to see what you can build, then let AI fill it in. Every template is a real, editable starting point on the same infinite canvas.
Mind Map
Branch one central idea out into themes and sub-points, then drag the connections around until the structure makes sense.
What is a knowledge graph?
A knowledge graph is information stored as a network rather than as a hierarchy. Instead of documents inside folders, you have entities as nodes, meaning people, concepts, sources, events, claims, and relationships as labeled edges between them: cites, contradicts, depends on, is an example of, was influenced by. Google popularised the term for the structure behind its search results, and the same idea in a personal or research context is the reason tools like Obsidian and Roam produce a graph view. The claim underneath it is straightforward: the useful part of what you know is usually the relationships, and a folder structure destroys exactly that.
The distinction that matters is between an automatic graph and a deliberate one. Most note tools draw a graph from links you happened to make while writing, which produces a picture that is fascinating to look at and hard to use, because every edge means the same thing, which is that two notes mention each other. A knowledge graph in the useful sense has typed edges. Knowing that a paper cites another paper is worth little; knowing that it contradicts it, or replicates it, or depends on its method, is the thing you actually wanted to record, and it is the thing an untyped link cannot express.
The second thing a graph does that a folder cannot is show absence. A concept with one source behind it looks identical to a concept with fifteen when both are files in a directory. Drawn as a network, the thin node is visibly thin, the cluster nobody has connected to the rest of the work is visibly isolated, and the claim your argument depends on that you have never found evidence for is a node with an empty side. Those are the findings people want from a knowledge base and almost never get, because search answers questions you thought to ask.
Storyflow is an AI-native infinite canvas, and it is deliberately the informal end of this. There is no schema to design before you begin, no query language, and no triple store. Describe what you are working on and the AI lays out the entities it can identify with candidate connections, then you correct it: fix the wrong edges, label what the relationships actually are, drag the clusters apart until the structure means something, and pin the underlying PDFs, quotes, and links to the nodes they belong to.
HOW IT WORKS
From a pile of notes to a structure you can read.
No schema, no query language, no ontology decided in advance.
01
Open a canvas
Sign up in seconds. No download. Your infinite graph canvas opens in the browser.
02
Describe the territory
Type the subject and what you are trying to work out. The AI lays out the entities it can identify with candidate connections between them, so you start from a drawn network.
03
Label what the edges actually are
Cites, contradicts, depends on, is an example of. A labeled edge is the unit of knowledge here, and correcting the AI's guesses is where most of the understanding happens.
04
Read it for what is missing
Find the thin nodes, the isolated clusters, and the claims with nothing behind them. Then share a view-only link or export the graph as an image or PDF.
Nodes, typed edges, sources, and the gaps.
A structure that answers questions you did not know to ask.
AI extracts entities and proposes the first connections
Start from a drawn network
Describe the subject or drop in your material and the AI lays out the entities with candidate edges between them. Correcting a proposed structure is far faster and far more instructive than building one node at a time.
See the AI mind map generator →
Every edge says what kind of relationship it is
Typed edges, not just lines
Contradicts, replicates, depends on, is an example of, was superseded by. An untyped graph is a picture; a typed one is a structure you can reason with.
Explore the concept map maker →
The source sits on the node it supports
Keep the evidence attached
Pin the PDF, the quote, the link, or the video frame to the node it justifies. A claim you can trace back to its source in one click is a claim you can still defend a year later.
See the research organizer →
Absence is visible as empty space
Find the thin node before someone else does
A concept resting on one source, a cluster connected to nothing, a claim with no evidence attached. These are the questions a search box will never surface for you.
Try the second brain →Unlimited boards. No object cap.
Open a canvas, describe the subject, and let AI lay out the first network. Storyflow is paid-only during early access, with plans from $7.99 a month billed annually, and a Free plan launches before the end of 2026.
Unlimited graphs on an infinite canvas with no node cap
Basic AI usage to extract entities and propose connections
Starter templates for research, mind maps, and second brains
Share view-only links with collaborators, no account required
BUILT FOR STRUCTURE
A folder tells you where a thing is. A graph tells you what it means.
Storage was never the hard part. The hard part is noticing what your knowledge is missing.
The edges carry the meaning
Type the relationship, not just the link
Support and contradiction: Two sources agreeing and two sources disagreeing look identical in an untyped graph, which is the single largest loss. Labeling them separately turns your notes into something you can argue from.
Dependency: Mark which claims rest on which. When one source turns out to be wrong, the graph shows you immediately what else you now have to revisit, which is otherwise a memory exercise you will fail.
Direction: Influence, citation, and causation all run one way. An arrow costs nothing to draw and preserves a distinction that a plain line destroys.
Time and supersession: Mark what replaced what. A graph that does not record supersession quietly keeps outdated understanding at the same weight as current understanding, which is worse than not recording it at all.
AI that reads the whole canvas as context
Ask the graph what you have missed
Extract entities from your material: Drop in notes and sources and the AI identifies the people, concepts, and claims and lays them out with candidate edges. It is a first pass to correct, not an answer.
Interrogate the structure: Ask which nodes rest on a single source, which two clusters have no connection despite being about the same thing, or what a missing edge between two concepts might be. The AI uses your whole board, so the answers are about your work.
It proposes, you decide: Storyflow structures and suggests; whether two ideas genuinely contradict each other is a judgment only you can make, and it is the judgment that makes the graph worth having.
Grow it without rebuilding it: Add a new source and ask the AI to place it. Nodes you have positioned and edges you have labeled stay exactly where they are.
Text, PDFs, images, and video on the same surface
Keep the raw material on the graph
Sources attached to nodes: Pin the paper, the interview transcript, the screenshot, or the dataset directly to the concept it supports, so the graph is navigable back to primary material rather than being a layer of abstraction over a folder you still have to search.
Quotes at the point of use: Put the actual sentence on the edge that claims a source supports something. Paraphrase drift is how a literature review ends up asserting things nobody wrote.
Video and web reference: Grab frames from YouTube or Vimeo, paste links and screenshots, and keep them beside the node they inform rather than in a bookmarks folder nobody revisits.
One graph a group can build together
Shared understanding, actually shared
Collaborators on one canvas: A research group, a reading group, or a team onboarding someone new can build one graph rather than several private ones that quietly disagree.
View-only links for readers: Send a link and anyone opens the whole graph in a browser with no account, which makes the structure of an argument something you can hand to a supervisor or a colleague directly.
Comments on the node in question: A disagreement about whether two findings actually conflict belongs on the edge between them, not in a separate thread that loses its referent within a week.
Explore templates for your project
Every template opens as a real, editable board on the infinite canvas. Pick the closest fit and make it your own.
Mind Map
Branch one central idea out into themes and sub-points, then drag the connections around until the structure makes sense.
HOW WE COMPARE
Storyflow vs Obsidian, Roam, and Neo4j.
These solve genuinely different versions of the problem: Obsidian is excellent for local, plain-text linked notes, Roam pioneered the outliner with bidirectional links, and Neo4j is a real graph database for querying at scale. Here is where they differ.
Storyflow
Recommended
Freeform spatial layout you arrange by hand
AI extracts entities and proposes connections
Labeled, typed relationships between nodes
PDFs, images, and video attached to nodes
Local plain-text files you own on disk
Query language for large-scale graph queries
Obsidian
Freeform spatial layout you arrange by hand
AI extracts entities and proposes connections
Labeled, typed relationships between nodes
PDFs, images, and video attached to nodes
Local plain-text files you own on disk
Query language for large-scale graph queries
Roam Research
Freeform spatial layout you arrange by hand
AI extracts entities and proposes connections
Labeled, typed relationships between nodes
PDFs, images, and video attached to nodes
Local plain-text files you own on disk
Query language for large-scale graph queries
Neo4j
Freeform spatial layout you arrange by hand
AI extracts entities and proposes connections
Labeled, typed relationships between nodes
PDFs, images, and video attached to nodes
Local plain-text files you own on disk
Query language for large-scale graph queries
What creators are saying
Join early creators getting structured workspaces and AI that remembers their projects
“Storyflow has sped up my workflow by at least 3x, which means more flow state and more projects I can actually ship. It truly changed the way me and my team create.”
Reilin Joey
Director & YouTuber
“One prompt gets me a structured board. But the tactics are my favorite. I run my YouTube scripts through them and my intros and retention got better. It's amazing.”
Justkay
YouTuber & Freelance Filmmaker
“I used to juggle five apps to plan a project. Now I describe what I am making and get boards, lists, and a schedule. All in one place.”
George
@fernwehchronicles
Knowledge graph questions, answered.
What one is, how to build a useful one, and where the limits are.
A knowledge graph stores information as a network of entities and the labeled relationships between them, rather than as documents inside folders. The nodes are things: people, concepts, sources, events, claims. The edges say how each pair relates: cites, contradicts, depends on, is an example of, was influenced by. Google popularised the term for the structure behind its search results, and the same idea applied to personal or research knowledge is why several note tools offer a graph view. The argument for it is that the valuable part of what you know is usually relational, and a folder hierarchy forces every item into exactly one place while discarding everything about how it connects to the rest.
Find out what your knowledge is missing.
Open a canvas, describe the subject, and let AI lay out the first network. No download.