ZettelFlow

Most plugins help you write. Almost none help you think.

Stop managing notes. Start managing knowledge. Obsidian is the store; ZettelFlow is the engine that makes the knowledge inside it evolve — it doesn't manage notes, it manages the processes that transform knowledge. It stands on four pillars:

• 🌱 Knowledge Lifecycle — every idea has a state (fleeting → permanent → evergreen); ZettelFlow knows the phase and proposes the next move.
• 🕸️ Semantic Knowledge Graph — relations with meaning: ask "show me every idea that contradicts this" and follow reasoning chains.
• 🔭 Knowledge Discovery — surface the unexpected: connections you didn't know you'd made, ideas worth challenging, a book outline hiding in a folder.
• 🩺 Knowledge Health — measure balance and richness, not volume: maturity, knowledge debt, a weekly review, a heatmap of ideas developed.

AI is one Action, never the product — every pillar works fully with AI disabled, and when it is on it proposes, never commits without your verdict: you accept, edit or reject every suggestion. Preserve people's work; simplify the paths: the canvas wizard you know remains the engine's front end while overlapping interfaces are consolidated with compatible access.

Our direction: useful from the first idea, deeper as your practice grows. You choose what deserves attention; recorded activity is not a measure of your understanding. The proposed Knowledge with purpose epic tracks purpose-led Cultivate and first value on your own material. The implementation is described below; real-device walkthroughs and the consented adoption pilot remain pending, not proven by automated tests.

Read the manifesto → · full documentation →

---

How it works, in one line

ZettelFlow turns an Obsidian Canvas into a guided note-creation wizard. Draw your workflow as a graph — steps as nodes, order as arrows — and the plugin walks you through it every time you create a note, filling in frontmatter, body content, dates, selectors, and more without you touching a template manually. Each step is a piece of the cognitive engine; each note lands already related, cross-checked and scored.

---

Zettelkasten toolkit

ZettelFlow is more than a note templater — it is a full Zettelkasten workflow. Beyond the canvas wizard, it ships a set of tools that grow and maintain your slip-box:

• 🎯 Knowledge with purpose — start with your own note or question, inspect bounded recorded context, write a provisional response, and save/pause/resume one inquiry locally. Evidence does not close your question: only you declare it sufficient for now. Markdown snapshots cite only material you explicitly marked as used and never overwrite an existing note. No AI or workflow setup needed; ordinary Cultivate remains available. See Cultivate.

• 🏠 ZettelFlow Home — open ZettelFlow, not Obsidian. A single front door: a greeting and "you've been thinking for N days", a one-click 3D-graph teaser, a growth nudge ("N fleeting notes ready to develop") that jumps straight to the latest capture, a Cultivate on-ramp (act on your next idea), your new ideas, main concepts, notes that deserve a review, suggested connections, and a "what to do next" list of prioritized recommendations (click to jump to the note). Read-only, offline — the state of your mind at a glance.

• ⚡ Quick capture — from thought to note in one keystroke. A command opens a single title prompt and writes a fleeting note to your Inbox — no canvas, no wizard, the fastest path to a first note. Home then nudges you to develop it later. Works on mobile; bind your own hotkey.

• 🌱 Cultivate — make one idea evolve. A guided thinking session picks your highest-leverage idea and walks you through small, real moves on it — connect an unlinked neighbour, challenge it with its contradiction, capture a question, advance its lifecycle state, add a source — each a one-click operation on the note. The session refines live as you go, and the metric (degree, maturity, state) is a consequence of the work. Offline; AI is never required. This is the engine that makes knowledge evolve, not just get stored.

• ⚖️ Agency review — see how much you're still deciding. A read-only Health → Agency tab that lists your recorded decisions newest-first with a compact header: the cognitive agency index and the accept/modify/reject breakdown of AI/derived suggestions, plus a one-line plain-language reading. Local, offline, nothing transmitted — a description of your verdict mix, never a grade.

• 🖼️ Shareable idea card — show how your thinking grew. One click on the Evolution timeline paints a before→after image of a note — first vs current state, claims gained, links, decisions and days — and saves it as a PNG to share. Built only from already-recorded data; nothing invented, nothing sent.

• 🔎 Ask your graph — query your thinking, not your frontmatter. A deterministic query surface over the meaning and structure of your slip-box: compose predicates — state:permanent, relation:contradicts, degree>=5, orphan, unsourced, older-than:30 — with AND/OR to answer questions Dataview can't, like "orphaned permanents older than 30 days that contradict X" — or compose one from menus with the guided term builder if you'd rather not memorise the grammar. Every result opens on click; view them as a list or a table; save a useful query — name it, reorder it, and pin it to Home as a live "N notes match …" card. Never AI, always offline.

• 📝 Note-builder companion pane — a live preview of the note as you build it, plus suggested connections to existing notes so you can link before you file.

• 📖 Derived projects — turn a folder of notes into the structure of a book / course / article: one command clusters and orders them from the semantic graph into an outline (MOC) that links every source note. It organizes what you already know — graph-derived, offline, no AI.

• 🔑 Zettel ID action — stamp every note with a stable unique identifier (sortable timestamp or Luhmann-style Folgezettel branching like 21 → 21a → 21a1).

One front door, four surfaces. The panels below live in four surfaces — Home, Health, Discovery and Graph — each with modes behind a segmented control, opened from a single ribbon button and as normal Obsidian tabs. Nothing is lost; every panel is a mode. See The four surfaces.

• 🩺 Slip-box health — one home for the state of your system. Connectivity % (connected / orphaned / unresolved) and a "today" panel (to process · contradictions · connections · open questions, each with a next action); a Knowledge Debt score (0–100) with a severity bar and a per-category drill-down (unreferenced · dangling · unsourced · open questions) each one click from a fix; a Knowledge balance read-out (references · questions · examples · conclusions · concepts) with nudges; and the orphan / dead-end lists. Read-only, offline — so knowledge-debt never piles up unseen. (The former standalone "dashboard" is folded in here.)
• 🧩 Systems Gallery — somewhere to start. Install a complete knowledge system in one click from the community browser: pick a folder and ZettelFlow writes a ready-to-run canvas plus its step notes (no clipboard, no manual paste), then Run now to use it immediately. Every system carries a difficulty badge (easy · medium · hard) so you can start where you're comfortable and grow into the richer ones. Start with the 🎓 ZettelFlow tour — a three-step guided system that teaches the whole workflow while you build a real note. A dozen shipped systems — Academic research · Zettelkasten v2 · PARA v2 · GTD · Reading · Writing · Software architecture KB · Meeting notes · Daily journal · Concept note · Decision journal · Inquiry — each composing the cognitive actions on creation (the medium/hard ones wire up the full relation · research · maturity pipeline), so a new note lands already related, cross-checked and scored. Offline, no AI. This is the one way to adopt a workflow — it replaces the old starter-flow and methodology-package installers.
• 🗺️ Map-of-content builder — gather notes by tag/folder into a MOC and refresh it safely; re-runs never touch your own prose.
• ✨ Connection resurfacing — "talk to your slip-box": for the note you're reading, see older, related notes worth revisiting, plus a daily spark of forgotten ideas.
• 🗓️ Second-brain review — one command generates a weekly review note: ideas created, orphans, forgotten ideas, and important-but-unreviewed notes over the last 7 days — each a click from a fix.
• 🔥 Thinking heatmap — a GitHub-style calendar of ideas developed (state advanced, source or connection added) over the last year — momentum, not note-count volume. Fed by a private, local, on-by-default journal (day → count only; no content, no network).
• 🔭 Morning discovery — up to three surprising connections: pairs of notes that share concepts but aren't linked yet, one click from relating them. Graph-structural, offline — the value of a slip-box is in the links you didn't already know about.
• 🗺️ Living knowledge map — detects your hubs and the notes that orbit each one, and regenerates as the vault changes so it never goes stale. Read-only, offline — the shape of your knowledge at a glance.
• 🕸️ Concept navigation — walk your vault like a wiki you wrote: focus a note, see its typed neighbours (in and out), click one to re-focus — Learning → Memory → Spacing effect → Anki, no folders. Ships with headless reasoning paths that read the graph as an argument (supports → expands → example → implements). Read-only, offline.
• 🌌 3D knowledge graph — see the shape of your thinking. Fly through an immersive Knowledge Galaxy: your slip-box as a 3D force-directed graph over a starfield, notes sized by connectivity, cluster-hued glow halos, links coloured by relation type; search-to-focus, filter by state, and a discovery lens that lights up orphans, dead-ends and contradictions in space. Click a node to open it, take a cinematic tour through your hubs, or export the view to a PNG / WebM clip to share. Read-only, offline; respects reduced-motion, and degrades to the 2D map on mobile. (Graph surface → 3D mode.)
• ❓ Open questions — every unanswered question in your vault, made first-class: each question:: with no answer yet, its asker(s), and the note most likely to answer it (ranked by shared graph context). Read-only, offline — a live thread instead of a dead end.
• 🕰️ Evolution timeline — the conceptual history of an idea (not a text diff): a per-note sequence of its lifecycle state + claims, captured only on meaningful change. Local, bounded, opt-in (it stores note content) — see how a claim evolved from "AI replaces us" to "AI is a copilot".
• ⚖️ Evidence map (experimental) — ask "what do you think about this?" and get a grounded synthesis reconstructed only from your graph: what supports it, what contradicts it, the sourced evidence, the gaps — every row linked to the source note, no invented content, no AI required.
• ✂️ Atomicity split assist — turn a multi-topic note into linked atomic notes in one command, leaving the source as a hub.
• 🌱 Note lifecycle states — give every note a state (🌱 fleeting → 📝 literature → 💡 permanent → 🔬 developing → 📚 evergreen → 🪦 archived) and move it with a validated Change note state command; the state lives in plain, configurable frontmatter.
• 🧭 Knowledge phases for steps — tag each step in a flow with the stage of knowledge work it advances (Capture → Classify → Process → Connect → Develop → Review → Consolidate); the step builder groups your flow by phase, so a workflow reads as an arc of thinking. Distinct from a note's lifecycle state: a step has a phase, a note has a state.
• ⚡ Event-driven workflows — let a flow react to vault events (a note created, a property or tag change) instead of only running on demand. Opt-in and off by default, throttled, and loop-guarded so a flow can never retrigger itself; an optional zf condition gates each trigger.
• 🧩 Visual workflow language (WHEN / IF / ACTION / WAIT) — compose a reactive flow on the canvas as a readable program: WHEN a vault event happens, IF a condition holds, run an ACTION, then WAIT for your confirmation. WAIT is a human-in-the-loop pause; the blocks are annotated on the canvas so a flow reads as an arc of thinking.
• 🧠 Knowledge actions — actions that reason about your slip-box, not just write to a note: detect an orphan, calculate maturity (0–100 from state, links, sources and age), find contradictions, find unanswered questions, suggest the next move (a concrete to-do: add a source, connect it, add an example, develop it), and run the thinking simulator (critical-thinking prompts adapted to a note's gaps — what if it's false? what evidence is missing? what contradicts it?) — deterministic, offline, over the knowledge model. They feed the knowledge-health layer.
• 🔗 Relation actions — actions that turn connection-making into a workflow step: find related and suggest link rank notes worth linking by shared graph context (co-citation + coupling), and create semantic relation writes a typed edge (supports, contradicts, …) to a target note. Graph-structural, deterministic, offline. To undo one, the Remove a relation command lists the active note's typed relations and deletes the one you pick, after a confirmation.
• 🔍 Research actions — actions that make sourcing and claim-checking part of the workflow: extract claims, compare claims (surface notes that agree with or contradict yours), find sources (suggest existing vault sources for an under-sourced note), and attach source. Over the claims-and-sources model, deterministic and offline.
• 🤖 AI actions — optional, off by default — AI is one action category, never the core, and it never writes for you — every completion is a proposal you accept, edit or reject. The whole plugin works fully with AI disabled. Opt in with your own OpenAI-compatible provider (endpoint + key + model — OpenAI, OpenRouter, LM Studio, Ollama…) to summarize, classify, and generate questions. No bundled key, no telemetry; the note content is sent only to the endpoint you configure. See AI provider setup.

New here? Open the community browser and install a system — it writes a ready-to-run canvas and its steps, and offers Run now. Everything else lives under Settings → ZettelFlow.

---

How it works

Canvas file  ──►  ZettelFlow wizard  ──►  Note in your vault
(your workflow)   (step-by-step UI)      (frontmatter + body merged)

1. Design — Create a .canvas file. Each node is a step; arrows define the order. Mark one node as the root.
2. Configure — Right-click any canvas node → Edit ZettelFlow step to add actions (prompt, calendar, selector, tags…).
3. Run — Click the ZettelFlow ribbon button → Create note (or bind a hotkey to the Open workflow command). ZettelFlow walks the canvas graph and builds the note.

---

Get started in 5 minutes

Fastest path:

1. Install ZettelFlow from the Obsidian community plugin browser.
2. Click the ZettelFlow ribbon button (or run Open ZettelFlow from the command palette).
3. On the welcome screen, click Start with my own idea, then choose an existing note or capture one.
4. State a question, inspect selected material, write a response or gap, and Save progress. Optionally install a system from the existing gallery when you want a Canvas workflow.

Manual path (build your own flow):

1. Create a .canvas file (e.g. flows/daily-note.canvas).
2. In Settings → ZettelFlow, set that canvas as the "new notes canvas".
3. Add a note file to the canvas, right-click it → Create managed step, enable Root.
4. Click the ribbon button → Create note — your first wizard run.

Stuck? Read the Getting started guide or open a Discussion.

---

Features

---

Capabilities & privacy

ZettelFlow collects no telemetry and sends no personal data or vault contents anywhere. The plugin uses these capabilities:

• File system (vault). Reads canvas flow files and creates/edits notes (e.g. the Change note state command writes a single lifecycle property to the active note). All access goes through Obsidian's Vault / FrontmatterService API — never a hardcoded path. Works on desktop and mobile.
• Vault enumeration. ZettelFlow lists every markdown file path to build the offline knowledge model that health, discovery, the graph and Cultivate all read. Paths and metadata only, all local — and you can keep folders out of it with the knowledge scope.
• Network — two opt-in paths, nothing until you use them. The community gallery does read-only GETs of the static catalog on GitHub (no backend, no account, no uploads), and the optional AI provider sends length-bounded note content to the single https endpoint you configure. Both are off until you open the browser or enable AI.
• Dynamic code execution. The Script action, dynamic selectors, vault hooks and workflow-event conditions run JavaScript you write, with the plugin's access to your vault — including read-only access to the whole knowledge model via zf.knowledge — so only run scripts you trust. No remote code is ever fetched or executed, and every runtime function is built in one audited module.
• Clipboard — write only. The “copy” buttons put a step or action configuration on your clipboard as JSON. ZettelFlow never reads your clipboard.

See Capabilities & privacy for full details.

---

Contributing

• Bug? → Open a bug report
• Idea? → Open a feature request
• Question? → Start a discussion
• Code? → Read Contributing & conventions and open a PR.

---

Support

If ZettelFlow saves you time, consider supporting development: