Weave

Weave is flowss's drag-and-drop diagramming canvas: drop shapes from a library of 271, connect them by dragging between handles, and arrange them with smart guides, auto-layout and frames. It is also more than a drawing surface. A Weave board knows its notation (flowchart, swimlane process, ERD, C4, sequence, org chart, mind map or network), checks itself for dangling connectors and unreachable steps, can be run as a flow simulation that shows where work piles up, can be scheduled with a critical path, carries typed data on every shape, and round-trips to Mermaid, DOT, D2 and DBML text. This page is the complete reference. Open Weave at /lucidflow.

At a glance

Who can use Weave

Weave is one of the seven studios that open on Starter and above; Free accounts reach the Studio and Sketch only. If your plan cannot open Weave, visiting /lucidflow takes you to the pricing page with Weave marked as locked. A signed-out visit goes to the sign-in page and returns you to Weave afterwards, keeping any template or shared diagram in the link.

Everything on the canvas works without AI. The features that call a language model are the flowss Weave Agent (Quick mode of the prompt, and its multi-step Visual designer role in Agent mode), Generate ideas from this branch on a mind map, the Deep analysis in Describe canvas, and recognising a diagram from a photo through Import a photo, paste or link. They work on Starter with your own AI key (Bring your own AI key) or on Plus and above with hosted AI points (AI points and limits).

The workspace

The canvas fills the page, and every control floats on it in one of five places.

When you select something, a selection toolbar appears beside it. The Inspector drawer opens on the right when a panel needs room (⌘. opens and closes it); it can be pinned open.

The first time you open Weave, the board holds a small starter flow (Start, Refine, a decision "OK?", Approve and Reject). An empty board shows "Nothing on the board yet." with Browse templates, Add shapes and Type a flow.

The studio menu

Selecting Weave (the flowss mark and the studio name, top left) opens the studio menu, the same menu every flowss studio has. While a work session is under way, a small dot in the phase's colour sits on the mark, and hovering it reads, for example, "Weave menu · Flow: In flow".

The ⋮ file menu

The ⋮ button beside the title (its accessible name is "File") holds the board's document commands, in four groups:

The commands in these menus are also in ⌘K under the same names.

The view island

Hide interface (⌥Z, or two fingers held still on the canvas on a touch screen) puts every island away so the canvas has the whole window; ⌥Z, the same hold, or Escape brings them back.

The tool rail

The status and readiness

The save status beside the title shows the board's size (shapes and connectors), any check errors and warnings, and a Diagram checks reading ("Clean", "Not checked" or counts) that opens the checks when selected. A readiness chip uses the platform's shared words, Empty, Draft, Sound, Reviewed and Publishable, so you can see at a glance where the board stands (see The work model).

If the browser refuses to save (storage full, or site data blocked), the status turns to "Not saved — storage refused" and explains: "Your board is still in this tab, but it will not survive a reload." Use Export a copy at once. If only the version history is refused, the status says "versions paused"; the board itself is still saved.

Adding shapes

The library

Shapes and Icons on the rail open the library flyout, which has three tabs, Shapes, Icons and Templates, and can be pinned open.

On the Shapes tab, Search shapes filters every group by name. Click a tile to add the shape in the middle of the view, drag it to place it, or focus it with Tab and press Enter. You can also right-click the canvas and choose Add shape here. The 46 groups are:

Basic holds Rectangle, Rounded, Stadium, Ellipse, Diamond, Triangle, Pentagon, Hexagon, Octagon, Parallelogram and Trapezoid. Callouts & containers holds Callout, Container, Swimlane, Pool and Sticky note. Tables & time holds the interactive objects described in Interactive objects.

Icons

The Icons tab has 34,766 built-in icons that work offline. Click or drag one in. Quick chips (arrow, user, server, database, cloud, device, lock, chart) start common searches, and Browse by category… lists the built-in icons by category. Type at least two characters in Search icons — aws, heart, kubernetes and a second section, Online · Iconify (200k+), adds matches from the online catalogue, fetched through flowss's own server so ad blockers and strict proxies do not empty it. If the online catalogue cannot be reached the section says "Couldn't reach the online icon catalogue — check your connection. The bundled icons above are still searchable." See The icon library.

Templates

The Templates tab (also ⋮ → Templates…) lists two kinds, and they behave in opposite ways, so the workspaces come first and look different:

The 47 workspaces appear in one list, in this order. Hover a card for its description.

The diagram templates number over 200. Among them are a cloud and infrastructure set (for example "AWS — Three-Tier Web App in a VPC", "Kubernetes Cluster Topology", "Zero-Trust Network Access", "Data Lakehouse (Medallion Architecture)" and "Disaster Recovery Topology (Pilot Light)") and a business set whose titles start "Weave —" (for example "Weave — Order to Cash (Swimlanes)", "Weave — Service Blueprint (Frontstage / Backstage)", "Weave — Procurement Approval Chain" and "Weave — Budget Approval Workflow"). The diagram list loads in the background after Weave opens and reads "Loading templates…" until it arrives; a filter that matches nothing says "No matches."

Loading a diagram template or starting a new mind map replaces the canvas. The board you were editing stays in your Library as it was, and ⌘Z brings it back to the canvas. Template links from /templates (/lucidflow?template=<id>) open the template directly. See The template gallery.

Creating from the keyboard

With exactly one ordinary shape selected:

With several shapes or nothing selected, Tab and Enter keep their usual meaning.

Type a flow

Type a flow (⇧⌘T, / on an empty canvas, ⌘K "Type a flow", or the first row of the ⋮ menu) turns a sentence into a laid-out flow in one undo step. Its grammar:

A name used twice is the same step, which is how you loop back. For example:

Start -> Review -> Approve? yes: Ship, no: Fix -> Review

Use the example fills that in. As you type, the dialog shows how many shapes and connectors it will make and the Mermaid it reads, and lists any errors by line. A document that starts with flowchart or graph is read as Mermaid as it is. ⌘Enter (or Insert) adds it; Esc or Cancel closes.

Connecting shapes

Hover a shape to show its handles and drag from any handle to another shape to connect them. Double-click a connector to edit its label. Select a connector and the selection toolbar offers:

⌘K also has Change connector kind, Label the source end and Label the target end (the labels beside each end, where an ERD reader looks for "1" and "0..*").

Line jumps at crossings (⋮ → Arrange, or ⌘K Connector line jumps) makes one connector hop the other where two cross. The style (None, Arc, Gap or Square) is chosen in Layout options. Jumps are saved with the board, so exports show them too.

Selecting and editing shapes

Click to select; Shift-, ⌘- or Ctrl-click adds to the selection; the Box-select tool or dragging a marquee selects an area; ⌘A selects all. Double-click a shape (or press F2) to edit its label. While you drag, smart guides align the shape with its neighbours; otherwise shapes sit on an 8-pixel grid. The corner handle rotates a shape ("Drag to rotate · hold Shift for 15° steps").

The selection toolbar for shapes

⌘K adds Match width, Match height and Match size (each takes the first selected shape's size, resized about each shape's own centre).

The right-click menu on a shape offers Copy, Cut, Paste, Duplicate, Edit label, Bring to front, Send to back, Lock / unlock, Move to layer (once the board has layers, one row per layer), Align (and distribute, with three or more shapes), Auto-layout, Reshape as mind map and Delete; on the empty canvas it offers Paste, Select all, Add shape here, Auto-layout, New mind map and Fit to view; on a frame it adds Tidy into a grid; on a table it adds Add row, Add column, Remove last row and Remove last column.

Locking

A locked shape can be selected and read but refuses drag, resize, rotate, delete and new connectors, and align, distribute, match size, flip and auto-layout step over it. Its tooltip says "Locked — unlock to move, resize or delete (⌘⇧L)". A whole layer can also be locked.

Frames, groups, swimlanes and pools

Containers, swimlanes, pools and groups are frames: they sit behind other shapes and own whatever lies inside them, so dragging a frame takes its contents with it and duplicating it duplicates them. Membership is simply what is inside the frame's box; there is no separate "add to frame" step.

Mind maps

Mind map on the rail (or ⌘K New mind map) replaces the canvas with a starter tree that grows right in depth-aligned columns. With a mind node selected, Tab adds a child branch and Enter adds a branch alongside (these take priority over the flow keys). Each node has buttons to add a child, fold or unfold its branch (the caret), set a Branch colour (recolours the whole limb), Switch the map's direction, Comment on this branch, Delete this branch and Generate ideas from this branch (AI; a toast reports "Grew N ideas from …" and ⌘Z removes them). Reshape as mind map (selection toolbar or ⌘K) re-lays connected shapes as a tidy tree from their root.

Clipboard

⌘C, ⌘X and ⌘V copy, cut and paste shapes (pasted shapes are offset and selected), including between Weave tabs. Pasting text or a file that is not a Weave selection hands it to the import flow.

Arranging the board

Layout options lets you choose:

Apply runs it; Cancel closes. Shapes stay in their containers and lanes whichever scope you choose, and locked shapes are left where they are.

Notations and the Model inspector

Every board has a notation, which decides the roles shapes can play, the connector kinds offered, the checks that apply and the import and export formats available. Weave infers the notation from what is on the board (pools and lanes make a swimlane process, entities make an ERD, lifelines make a sequence), or you can set it.

A notation's import formats appear under ⋮ → Import and its export formats under ⋮ → Export only while that notation is active.

The Model tab of the Inspector (⌘K Model inspector) shows:

Diagram checks

Diagram checks (Analyse menu, the status reading, or ⌘K Check this diagram) names every problem rather than counting it, grouped by severity (errors, warnings, hints). Select a finding to select the shapes it is about. Findings that can be repaired mechanically carry a repair button; Fix all safe (N) (or ⌘K Fix all safe checks) applies every safe repair as one undo step: nudging overlaps apart, labelling a two-way branch, aligning a shape with its neighbours, moving a step into its lane. Nothing is deleted and nothing is re-pointed. Ignore dismisses a finding on this board, with an optional reason; ⌘K Ignore the checks on the selection does it for everything selected. Ignored findings are listed under Suppressed with a button to report them again. When there is nothing to report the panel says "Nothing to report. Every connector lands somewhere, every step is reachable, and no two shapes are sitting on top of each other."

Rule packs chooses which families run:

At least one pack always stays on: the last one left switched on cannot be turned off. Ignored findings and pack choices are saved with the board. On very large boards the checks stop at their limits and say so ("Board too large to lint in full") rather than freezing.

Flow simulation

Flow simulation (Analyse menu, ⌘K Run the flow) runs the diagram as a process. Tokens enter where work starts (shapes nothing points at), queue where a step is busy, take each branch out of a decision and leave where the flow ends (shapes pointing nowhere). No preparation is needed: an unannotated box is a task that takes one time unit.

The transport at the bottom of the canvas has play and pause, Restart, a scrubber and a speed menu (0.5×, 1×, 2×, 4×). Its header shows how many items went through, the mean lead time and the throughput per time unit, for example "24 through · lead time 6.5 hour · 1.2/hour" (the unit is printed in the singular), followed by Bottleneck: and the step where work piles up most. A line explains where the timings came from: "Timings: 3 from shape data, 4 read from labels, 2 assumed to take one hour." Steps holding work wear a halo whose size is the queue. The simulation never changes your board.

To sharpen the answer, give steps durations in either of two places:

Shape data wins over the label. The time unit is chosen so the numbers stay readable.

Critical path

Critical path (Analyse menu, or ⌘K) turns the board into a schedule, using the same durations as the simulation. It shows the project duration ("12 days end to end"), how many critical paths there are, the Chain: of critical tasks (select it to select them on the board), and Shorten these to move the date: the top four tasks ranked by how much shortening them would move the finish, with "one of N paths" where a task is on only some critical paths. It handles several starts and ends and links with lag.

If the board has a loop, the panel names it and explains that a schedule needs a direction: break the loop for scheduling, or model the rework as its own step. Up to 5,000 tasks are analysed; problems are listed (two in full, then a count).

Shape data, conditional formatting and linked data

Shape data (Analyse menu, the Inspector's Data tab, or ⌘K) turns a diagram into a small database. It has three tabs.

Link data (Data tab, or ⌘K Link data (CSV / JSON)): paste rows or Open a file (.csv, .tsv or .json, up to 2,000 rows), choose the Key column, and choose what it is Matched to: the shape's label or a field it carries. Matching ignores surrounding spaces and letter case, so "Review " finds "review". Before anything is written the preview shows "N of M rows match a shape", lists "No shape for: …" (rows that matched nothing), "Repeated key, last row wins: …" and rows with no key that were skipped. Link writes the values onto the shapes. Link to Sheet node reads a spreadsheet already on the canvas instead, and stays live: edit a cell and the shape it feeds repaints. Each link in the list can also draw a Graphic on its shapes: pick one of the linked fields (or none) and whether it is drawn as a bar or a mark.

Export shape data (CSV) writes every shape's fields with its id first; re-import it keyed on id and it lands back on the same shapes. Limits: 100 rules, 50 data links, 128 fields and 5,000 records per board.

Interactive objects

The Tables & time group holds objects you work inside:

In a Sheet, a circular reference shows #CIRC!, and the volatile functions TODAY, NOW, RAND and RANDBETWEEN are not available (they show #NAME?).

Layers and the outline

Layers and outline on the rail opens a panel with two tabs.

Layers are sheets of the board, each with an eye and a padlock. Add a layer, rename it, give it a colour, move it up or down, delete it, and use the arrow on a layer row to move the selected shapes there (⌘K Move the selection to a layer). A hidden layer is left out of the checks, the description, the simulation and every export, and the Export menu says so ("omitting N hidden layers"). A locked layer's shapes cannot be moved or edited. Up to 64 layers.

Outline lists every shape in reading order with its lane, connectors and comments. Enter on a row focuses that shape on the canvas. It is the board as a document, for keyboard and screen-reader users.

The Source panel: the board as text

Source (Inspector tab, the ⋮ menu, or ⌘K) shows the board as text in your choice of Mermaid, DOT, D2, PlantUML or DBML (plus the active notation's own formats, such as SQL DDL, Structurizr, Markdown outline or CSV). Edit the text and Apply to canvas: a new line adds a shape, a renamed shape is renamed on the canvas, and everything the text still names stays exactly where you placed it. Revert to the canvas discards your edits, the copy button copies the text and Import text opens or pastes text from elsewhere.

Mermaid, DOT, D2 and DBML can be read back; PlantUML and the other written-only formats say "… is written, not read". A board too big to write as text says so and how many shapes were left out.

⌘K also copies the board straight to the clipboard: Copy as Mermaid (containers become subgraphs), Copy as DOT (clusters), Copy as D2 and Copy as PlantUML (packages).

Describe canvas

Describe canvas (the ⋮ menu, or ⌘K) opens What this diagram says: a deterministic paragraph and alt text built from the shapes and arrows ("the same canvas always produces the same paragraph"), with Copy prose and Copy alt text. Deep analysis → Analyse (AI) adds a structural reading, a reading path and observations, each checked against the board; anything it names that is not on the board is struck. ⌘K Copy the board's description copies the lanes, their members and the reading order, the same words an exported SVG carries.

Presenting

Walkthroughs

Present (top right, or ⌘K) starts a walkthrough:

The walkthrough controls show Previous step, "Step N of M", Next step, the step's talktrack, Autoplay the presentation and Present fullscreen. Keys: → and ← step, F goes full screen, Space pauses or resumes autoplay, Escape exits. A walkthrough needs at least two shapes.

Saved presentations

The Presentations tab of the Inspector keeps named shows in the document. Make one From selection, From slides or From the flow (⌘K Save a presentation saves the selection). For each step you can reorder it, remove it, write a talktrack ("what you say on this step"), set its Dwell time in seconds for autoplay, and Set camera to current view to frame the step by hand. Present a show with its play button. Autoplay advances on each step's dwell and stops at the last step. Up to 32 presentations and 500 steps.

Export PowerPoint (.pptx) makes each slide or presentation step a real slide: an editable title, the shapes inside it as bullets in reading order, and the talktrack as speaker notes. Export Word outline (.docx) writes the same deck as a document, one page per slide.

Version history

Weave takes a snapshot a few seconds after your edits settle, keeping up to 100 in this browser. The Versions tab of the Inspector (⌘K Version history) lists them with their time and author.

Comments

Pick the Comment tool and click anywhere to pin a thread ("Leave a comment — it pins right here and saves with the diagram…"), or use Comment on the selection toolbar to pin it to a shape. Threads take replies, can be resolved and re-opened, and can be deleted (deleting a thread removes every reply). Comments in the people island lists every pinned thread; select one to jump to its pin. Up to 200 threads, 2,000 characters per comment; a longer comment or reply is trimmed and a toast says so ("Comment trimmed to 2,000 characters"). Comments are part of the board and travel in a .weave.json bundle. Clear canvas deletes them permanently; ⌘Z after a clear restores the drawing only. See Comments and review.

Sharing and live collaboration

Share in the people island offers:

The seat belongs to whoever starts the session: guests on any plan, including Free, can join a session a Plus or Ultra member started, up to that host's limit. Voice and screen sharing are judged per person.

In a live session everyone sees each other's cursors and edits, and comments flow to everyone. Shapes too large to send live stay in your copy only, and a toast says so. If the connection drops, Weave keeps retrying and tells you when everyone's changes are flowing again. An invite link works for anyone who has it, so if one leaks, leave the session (your board is unchanged) and start a fresh one with a new link. Live presence, voice and screen sharing are temporarily unavailable in private projects ("Saved projects and comments still work"). See Live collaboration and Sharing and permissions.

Embed (Embed — iframe / Markdown snippet in ⌘K) opens Embed this flow, showing the embed link's size, with Copy link, Copy iframe and Copy Markdown, for Notion, Confluence and READMEs. See Embedding and the render API.

Import

The ⋮ → Import menu:

A BPMN 2.0 file lands with pools, lanes, tasks, gateways and events where the file's diagram places them. Anything the file does not draw arrives at a default size beside the rest.

You can also drag a text diagram file straight onto the canvas: .mmd, .dot, .d2, .dbml, .sql or .txt, up to 4 MB. It is read exactly as Import diagram text reads it, so a dropped PlantUML (.puml) file is refused even though the error message lists it; a dropped draw.io or .xml file is also read as text and refused, so use Import draw.io for those. Other files (pictures, for example) are refused with "Weave can't open a .png file here — try .mmd, .dot, .d2, .puml, .dbml or .sql"; bring pictures in through Import a photo, paste or link instead.

Export

The ⋮ → Export menu:

Hidden layers are left out of every export. Open in BPMN studio first shows a check of what will travel (participants, lanes, sequence and message flows) and anything that cannot be mapped. See Exporting your work.

The AI prompt

The prompt island at the bottom (⌘/ to focus it) has two modes:

The paperclip opens the import flow. Both modes need AI on your plan. See AI in flowss.

The flowss Weave Agent pane

Choosing Agent (or ⌘K Open flowss Weave Agent — Visual designer) opens the agent's pane beside the canvas, headed flowss Weave Agent, with its Visual designer role named beneath. ⌘K Open flowss Weave Agent opens the Quick prompt instead. It works step by step on the board, and its changes land on the canvas as it goes.

If a run ends early, the pane says so plainly: "The run stopped before it finished — its reply ended mid-run, so there is no result. Anything it had already applied is still on the canvas."

Preferences and view

Weave remembers which library tab, tool, connector style and side panel you last used.

Saving, the Library and Send to

Weave saves your board in this browser as you work, and the board appears in your Library, which syncs to your account. When the board is open from its Library document, select the title (or focus it and press F2) to rename it; Enter saves and Escape cancels. A board with no Library document behind it shows "Untitled flow" and its title cannot be edited in Weave; open it from your Library to name it. Clear canvas (⋮, then run it a second time to confirm) empties the board; ⌘Z brings the shapes back but not the comment threads. A board opened from the Library (/lucidflow?doc=…) opens in place.

Send to (in Share) sends the board to another studio, converting it where the target uses a different notation; Study receives it as course notes and as the source for diagram flashcards. See Moving work between studios and Cloud sync, offline work and devices.

The Flow dock and the Flow console

The Flow dock is flowss's continuity panel, the same in every studio. In Weave it opens from the small floating Flow pill (parked above the view island; drag it anywhere, double-click it to snap it home), from Flow in the studio menu, with ⌘J, or with ⌘K Open Flow Dock. The pill's ring takes the colour of your current phase.

Its heading carries your session's phase, read from how you have been working across studios in the last while. The phase word is a link to the Flow console, and, except while you are Idle, the line under it explains the reading and the move it suggests:

A streak of more than one day shows beside the phase (for example "3d").

The Flow console at /flow is the full-page version: where your work is, what it is part of and what comes next, across every studio. It reads only what this device already records and sends nothing anywhere. See The Flow layer and The continuum.

Keyboard shortcuts

On Windows and Linux, use Ctrl for ⌘ and Alt for ⌥. Press ? in Weave for its own sheet, Get fluent fast.

See Keyboard shortcuts.

Tips

• Write durations in labels, "Review (2d)", from the start: the simulation and the critical path read them with no extra setup.
• Use Type a flow for the first draft of any process, then refine by hand; it is faster than dragging five boxes.
• Set the notation explicitly in the Model inspector when Weave guesses wrong; the checks, connector kinds and export formats follow it.
• Run Fix all safe before presenting, and switch on the Geometry pack for the final polish.
• Put annotations on their own layer and hide it before exporting a clean picture.
• Name the versions you may want to come back to; unnamed snapshots stay in this browser only.
• Use Copy as Mermaid to drop a board into a README or pull request, and Source to edit a large flow as text.
• Keep a .weave.json bundle of any board that matters: it is the one file that carries comments, layers, presentations and named versions as well as the drawing.
• Start from a ready-wired workspace (Org chart, ER diagram, Swimlane process and the rest) when you want a real example beside your own work rather than in place of it.

Limits and known constraints

• Weave needs Starter or above; live sessions need the host on Plus (up to 3 people) or Ultra and Enterprise (no cap); AI features need your own key (Starter) or hosted AI points (Plus and above).
• The board is saved in this browser and synced through your Library; if the browser refuses storage, the board survives only until the tab closes.
• Unnamed version snapshots (up to 100) stay in this browser; only named versions travel with the document.
• Share links that carry the board are limited to about 1 MB of encoded data; larger boards (usually with imported images) must be shared live or exported.
• PlantUML and some notation formats can be written but not read back onto the canvas.
• Flow simulation durations: weeks and months are not read from labels.
• Critical path analyses up to 5,000 tasks and cannot schedule a board with a loop.
• Link data: up to 2,000 rows per inline paste or file; 100 rules, 50 links and 128 fields per board.
• Sheets: up to 1,024 rows by 256 columns; no volatile functions.
• Comments: up to 200 threads per board; 2,000 characters per comment.
• Layers: up to 64; presentations: up to 32 with 500 steps.
• A .weave.json bundle carries your 20 newest named versions, not the unnamed snapshots.
• Text files dropped on the canvas: up to 4 MB, and only diagram text Weave can read (.mmd, .dot, .d2, .dbml, .sql, .txt). PlantUML is offered by the import dialog but cannot be read onto the canvas yet.
• The Vault holds up to 40 checkpoints; a board too large to checkpoint (usually because of a pasted image) has to be exported instead.
• Voice in a live session: up to 8 people talking at once.
• The minimap is not available on phone-width screens.
• Live presence, voice and screen sharing are temporarily unavailable in private projects.

Troubleshooting

Related pages

• What flowss is
• A tour of the workspace
• Keyboard shortcuts
• Flow-systems analysis in Studio
• The icon library
• Importing and converting
• The template gallery
• BPMN
• Sketch
• Study
• Your library
• Version history and undo
• Moving work between studios
• Live collaboration
• Comments and review
• Exporting your work
• Embedding and the render API
• AI in flowss
• The Flow layer
• The continuum
• The work model
• Plans and what they include