Skip to content

Documentation

Make the most
of every idea.

Start with the essentials. Find the right engine, learn a new workflow, and get your best work out into the world.
Sculptural study of connected forms and structured ideas

Getting started

Start in Studio or on the Sketch canvas for free. Other workstations start on Starter.

  1. 1. Pick a workspace. StudioThe universal canvas — renders every engine.StudyInteractive knowledge maps, model-building and spaced retrieval.WeaveDrag-and-drop diagramming canvas.LabFlowScript — typed, composable diagram code.BPMNBusiness-process modelling (BPMN 2.0).ScienceBiology & science figure composer.FigureJournal-ready research figures from your dataset.SketchFreehand drawing — convert ink to diagrams.EvidenceTurn this diagram into an evidence scaffold to research.
  2. 2. Start from a template or describe what you want. The Templates browser has every starter point; the flowss Studio Agent will generate something from a sentence.
  3. 3. Refine in place. Click any label to rename it. Drag, recolour, resize. Ask the flowss Studio Agent for changes — “add an error branch”, “use a colour-blind safe palette”.
  4. 4. Export. PNG, SVG, PDF, TikZ, or the engine's native source. Share a permalink with the diagram encoded directly in the URL — the source travels in the link itself; whoever opens it signs in (free) to edit.

All documentation

Every studio, panel, engine, export, plan and setting, explained step by step — search it, or browse it shelf by shelf.

Studio

Code to diagram with every engine: the editor, templates, import, styling, icons and the flowss Studio Agent.

The Studio editorComplete reference to the Studio editor: figures, the code sheet, canvas navigation, click-to-edit, render errors, the file menu, undo, saving and shortcuts.Starting from a templateEvery way to start from one of 5,000+ templates: the Studio library, search rules, favourites, the public gallery, template pages and what loading replaces.Importing and converting (text, files, photos, data)Bring anything into the Studio: launchpad doors, paste, drag and drop, files, links, photos, Photo → Engine, SQL/DBML schemas, data and engine conversion.Styling, themes and layoutStyle Studio diagrams: layout direction, eight journal presets, canvas recolouring, engine styling syntax, light and dark themes, exports and workspace layout.The icon libraryBrowse, search, preview, copy and insert 34,766 offline icons from eight packs, plus the online Iconify catalogue, with licences explained.Data, timeline, TikZ and the diagram toolsData → Chart, the diagram timeline, TikZ/LaTeX export, click-to-edit diagram tools and the mind map outliner, control by control.Flow-systems analysis in StudioHow the Studio computes lead time, control limits, cut sets, queues and bottlenecks: every flow-systems engine and Diagram Intelligence.Choosing an engineA decision guide to 110 engines: which one for which job, close cousins compared, privacy, offline, conversion and plans.The flowss Studio Agent: prompt and quick helpEvery way to reach the flowss Studio Agent: the prompt in Quick or Agent mode, suggestion chips, dictation, canvas edits, the Fix buttons, explanations, costs and fixes.The flowss Studio Agent and its modesThe flowss Studio Agent's multi-step runs, and the named agents of the other studios: all 21 modes and pipelines, what each may change, approvals, conflicts, Revert and Undo, verification, limits and costs.

The studios

Study, Weave, Lab, BPMN, Science, Figure, Sketch and Evidence, each in depth.

StudyThe complete guide to Study: concept-map courses, lessons, spaced review, FSRS flashcards, the card browser, assessment, Scholar and the flowss Study Agent.WeaveThe complete guide to Weave: shapes, connectors, templates, notations, checks, flow simulation, critical path, shape data, live sessions and export.LabA full tour of the Lab: documents, the FlowScript editor, canvas, benches, side panes, the flowss Lab Agent, sharing, exports, shortcuts, limits and fixes.The FlowScript languageComplete FlowScript reference: syntax, values, units, uncertainty, expressions, checks, view lines, all 125 block kinds and every message.Lab views and renderersDeclare views in FlowScript, use the Figures gallery, and look up all 36 Lab renderers: what each is built from, computes and refuses.Reporting standardsScore Lab figures against 33 checklists such as CONSORT, PRISMA and STROBE, apply the repairs that ship, and start from compliant scaffolds.The Data and Statistics benchesBring rows into the Lab, clean and shape them, derive typed blocks, run eleven statistics calculators, and let the Advisor rank your next move.Publishing, provenance and compliance in LabTrace every number on a Lab figure, score it against reporting standards, preflight it for a journal and build a verifiable submission bundle.BPMNModel BPMN 2.0 processes with Camunda 8 wiring, linting and quick fixes, token simulation, event-log mining, the flowss BPMN Agent, handbooks, versions and live editing.ScienceCompose illustrated biology, chemistry, physics and maths figures whose reactions, units, constants and statistics are checked, run and pinned as you draw.FigureCompose journal-ready multi-panel figures from your data at the journal's real width, check them against its artwork rules, and export every format.Figure analysis and statisticsClean and reshape data, run 23 gated statistical analyses, draw fits and computed brackets, plan sample size, and copy a cited Methods paragraph.SketchThe flowss whiteboard: pens, shapes, binding arrows, frames and presenting, wireframes and workshop canvases, facilitation, and converting ink to diagrams.EvidenceTour of the Evidence studio: add papers, synthesize a quote-grounded evidence graph, read it, ask it questions, then save, share and export it.Screening, extraction and synthesisSystematic reviews in Evidence: RIS/BibTeX import, PRISMA screening and kappa, de-duplication, quote-grounded extraction, risk of bias, GRADE, living updates.

Engines

110 engines, one workspace. Each one renders in milliseconds and supports the studio's shared toolbar — click-to-edit labels, AI refinement, export, share.

Mermaid

compat: Mermaid 11

Versatile text-to-diagram language. Renders flowcharts, sequence, class, ER, state, gantt, mindmap, sankey, C4, packet, architecture, kanban, radar, treemap, venn, wardley, eventmodeling and 30+ more. Paste any Mermaid snippet — fully compatible.

Templates
344
Language
mermaid
Engine docsOpen ↗

Weave

compat: Lucidchart-style canvas

Full visual drag-and-drop canvas. Drag shapes from the palette, draw arrows between them, double-click to rename, move things around — exactly like Lucidchart, draw.io or Miro. The JSON source updates as you draw.

Templates
267
Language
json
Engine docsOpen ↗

Free-body diagram

Native free-body diagrams for mechanical, structural, and civil engineering. Declare bodies with `at x,y` positions and arbitrary force vectors with magnitudes, colours, and labels. Supports `moment` curved-arrow torques and an optional hatched `surface` line for incline / floor contexts. Renders in a true coordinate system so vectors are physically faithful.

Templates
85
Language
plaintext
Engine docsOpen ↗

Circuit schematic

Native electrical schematic capture. Declare components (resistor, capacitor, inductor, diode, LED, transistor, op-amp, source, ground, switch, junction) at grid coordinates with values, then wire them together. Renders with standard IEC / ANSI-style symbols and labelled values. Ideal for textbook circuits, lab notes, and small-design documentation; not a substitute for KiCad on a real PCB.

Templates
74
Language
plaintext
Engine docsOpen ↗

Drug-interaction network

Native pharmacology / clinical-pharmacy network diagrams. Declare drugs (with optional therapeutic class) and pairwise interactions with effect, severity, and mechanism. Edges are colour-coded by severity (minor → moderate → major → contraindicated) and stack cleanly when one drug pair has multiple interactions. Ideal for polypharmacy reviews and pharmacology teaching.

Templates
69
Language
plaintext
Engine docsOpen ↗

Digital logic (gates / K-map / truth table)

Native digital-logic engine — three modes in one. `type gates` for IEC-style gate schematics (AND, OR, NOT, NAND, NOR, XOR, XNOR, BUF), `type kmap` for 2-4 variable Karnaugh maps with Gray-coded rows / columns and highlighted 1-cells, `type truthtable` for full truth tables with output-coloured cells. Covers the EE / CS undergraduate digital-logic curriculum.

Templates
65
Language
plaintext
Engine docsOpen ↗

FlowScript

compat: flowss native — no other tool speaks it yet

The first diagram language designed for research. Every element has a domain type — `study`, `cohort`, `arm`, `outcome`, `project`, `task`, `service`, `kpi`, … — and numbers PROPAGATE. Declare `cohort eligible { n: 1240, from: screened, excluded: { ineligible: 198, declined: 64 } }` and the validator checks that 198 + 64 + 1240 = screened.n. One model renders as many publication-ready views (CONSORT 2010 flow ships in V0; Sankey, summary-bars, Gantt, C4, funnel coming next). Designed for clinical trials, biology, ML eval, project management, engineering architecture, and business strategy from day one.

Templates
62
Language
glyphscript
Engine docsOpen ↗

Causal DAG

Native causal directed-acyclic graphs for epidemiology and inference. Declare nodes by role — `exposure`, `outcome`, `confounder`, `mediator`, `unobserved` — and the engine colour-codes them and lays out the diagram so the causal story reads left-to-right. Bidirected (`---`) edges render as dashed double-headed arcs for unmeasured association.

Templates
57
Language
plaintext
Engine docsOpen ↗

Anatomy / body systems

Native anatomy diagrams for medicine, nursing, and biology teaching. Pick a body diagram (`whole`, `heart`, `lungs`, `kidney`, `brain`, `digestive`, `skeleton`) and drop labelled markers at specific anatomical points. Markers come with leader lines and an optional callout box so a single diagram supports both lecture annotations and exam-style labelling sheets.

Templates
57
Language
plaintext
Engine docsOpen ↗

Quadrant Matrix

Native 2×2 decision matrices and quadrant charts — Eisenhower, BCG growth-share, risk (likelihood × impact), effort/impact prioritisation, Gartner-style magic quadrants, stakeholder power/interest, and SWOT-as-scatter. Declare two axes, four quadrant labels, and `point` rows at 0..1 coordinates; points inherit their quadrant's accent or take an explicit `#hex`. Mermaid-quadrantChart-compatible syntax, rendered as crisp print-ready SVG.

Templates
55
Language
plaintext
Engine docsOpen ↗

Timeline (history)

Native historical / era timelines. Declare an axis with start and end years, drop labelled `event` points at specific dates, draw `era` bands spanning ranges (Bronze Age, Renaissance, Cold War), and stack lanes for parallel narratives (e.g. politics vs. science). Renders with smart label dodging so dense periods stay readable. Distinct from Gantt — built for centuries and millennia, not sprint days.

Templates
49
Language
plaintext
Engine docsOpen ↗

Genetics pedigree

Native genetics / genealogical pedigree charts using standard clinical notation: squares for males, circles for females, diamonds for unknown sex, filled symbols for affected individuals, half-filled for carriers, slashes for deceased, arrows for the proband. Mating lines link partners, sibship lines descend to children.

Templates
46
Language
plaintext
Engine docsOpen ↗

Petri Nets

compat: Native renderer

Concurrency's foundational model — places (circles) hold tokens, transitions (rectangles) fire when their inputs are saturated, arcs connect them. Tiny DSL: `place P1 tokens 2 "Ready"`, `transition T1 "fire"`, `P1 -> T1`. Auto-layout by topological depth, or pin every node with `@ x,y`. Native — no other major platform ships first-class Petri-net rendering.

Templates
45
Language
petri
Engine docsOpen ↗

Stratigraphy / Harris matrix

Native stratigraphic columns for geology and archaeology + Harris matrix for relative chronology. `type column` renders layers with thickness, colour, age labels, and depth ticks. `type harris` lays out context units in a DAG with `above` relationships, auto-levelling so each unit sits above whatever it physically overlies.

Templates
43
Language
plaintext
Engine docsOpen ↗

Phase diagram (materials)

Native phase diagrams for materials science and thermodynamics. `type pt` (pressure × temperature) renders a single-component diagram with SCHEMATIC sublimation / vaporisation / fusion curves and labelled triple and critical points — the curve shapes are qualitative and identical for every substance; the declared temperatures and pressures set the point markers and their captions, not the paths between them. `type binary` renders a composition × temperature alloy diagram against a numeric temperature axis read from the two declared melting points, with schematic liquidus and solidus curves bridging them.

Templates
43
Language
plaintext
Engine docsOpen ↗

State machine (FSM)

Native finite state machines, and the only engine here that decides whether the machine is one. A state machine's defects are facts about reachability and determinism, and neither has pixels: two arrows leaving a state on the same event with nothing to tell them apart is a machine and a coin toss, and it draws as two perfectly ordinary arrows. The engine reports nondeterministic branching, states nothing can reach, non-final states with no way out, a machine that can never terminate, and transitions naming states that were never declared — which every tool that draws what it is handed simply omits from the figure. Every arrow the document writes is drawn, and none is drawn through a box.

Templates
26
Language
plaintext
Engine docsOpen ↗

PERT / CPM network

Native project network diagrams with the critical path highlighted. Declare tasks with `duration` and `after`; the engine runs the full forward + backward pass and renders ES / EF / LS / LF / slack in every node. Zero-slack tasks are flagged as CRITICAL in amber. Finish-to-start only, in whole unit-less days — for a schedule on a real working calendar, with start-to-start, finish-to-finish and start-to-finish relationships, lead and lag, resources, deadlines and the compression ranking, use `gantt`, which draws the same network under `view network` from a schedule it also draws as bars.

Templates
23
Language
plaintext
Engine docsOpen ↗

Org chart

Native organisational charts from an indentation-based outline. One person or unit per line — `Name | Role | meta` — with 2-space indents for reporting lines. Auto-laid-out top-down tree of rounded cards with elbow connectors, depth-tinted accents, optional role and metadata rows (team size, location), and multiple roots for holding structures. The org chart Graphviz makes you fight for, in ten lines of text.

Templates
22
Language
plaintext
Engine docsOpen ↗

PRISMA 2020 flow

Native PRISMA 2020 systematic-review flow — and the only one that checks the arithmetic the standard is made of. A PRISMA figure exists so that every record entering a review is accounted for on its way out, which makes it an arithmetic claim: the engine balances all four stages (identification, screening, retrieval, eligibility), reports the exact count of records that vanished and where, draws the failing stage's arrow in red, and marks a stage it could not balance as a gap naming the missing term rather than passing it.

Templates
20
Language
plaintext
Engine docsOpen ↗

Reliability block diagram

Reliability block diagrams with the reliability actually computed: series, parallel, k-out-of-n and standby arrangements over blocks carrying R, MTBF or failure rate, evaluated to system reliability, MTTF and availability at a chosen mission time. Redundancy that buys nothing shows up as a branch that moves the system number by a rounding error — which is the entire point of drawing one.

Templates
15
Language
plaintext
Engine docsOpen ↗

Pareto chart

The 80/20 chart, computed: sorted bars with the cumulative percentage curve on a second axis, the 80% reference line, and the vital-few cut marked where the curve crosses it. Categories are ranked and any tail can be folded into an explicit `Other` bar rather than silently dropped, and a second period can be overlaid to show whether last quarter's vital few are still this quarter's.

Templates
14
Language
plaintext
Engine docsOpen ↗

Design structure matrix

Dependency structure matrices for architecture and process analysis. Declare elements and their dependencies and the engine renders the square matrix, then does the analysis a picture cannot: partitioning into a sequenced, block-triangular order, detection of the cyclic blocks that must be iterated, and counts of the feedback marks above the diagonal that quantify how much rework the current sequence forces.

Templates
14
Language
plaintext
Engine docsOpen ↗

Bowtie risk

The barrier-based risk diagram used in process safety, aviation and healthcare: threats on the left, the top event at the knot, consequences on the right, and preventive and recovery barriers on every path — each with its own effectiveness, escalation factors and degradation controls. Residual risk per path is computed from the barrier chain, so a decorative barrier is visible as one that changes no number.

Templates
14
Language
plaintext
Engine docsOpen ↗

Argument map

Native argument maps for philosophy, critical thinking, debate, and legal reasoning. Declare a central `claim`, then `support` and `object` arguments either attacking the claim or rebutting other arguments. PRO arguments render green, CON arguments red, with arrows pointing FROM each argument TO whatever it supports or attacks. Rebuttals nest naturally.

Templates
13
Language
plaintext
Engine docsOpen ↗

Value stream map

Native value-stream maps (VSM) for lean process analysis. Declare process boxes with their data — `process Weld { ct: 45s; co: 30m; uptime: 85%; operators: 2 }` — and the inventory or wait time between them, and the engine draws the timeline ladder and COMPUTES what the map exists to show: total lead time, total value-added time, flow efficiency, and the takt comparison. The one process diagram whose numbers are an analysis rather than a caption.

Templates
13
Language
plaintext
Engine docsOpen ↗

Flow metrics (CFD)

Cumulative flow diagrams with the flow metrics derived from the same data. Give it a date column and one cumulative count per workflow stage and it draws the stacked bands, then computes approximate WIP, average cycle time via Little's Law, throughput and flow efficiency, and marks where WIP ran away from throughput. The chart Kanban teams screenshot, with the four numbers they actually manage against.

Templates
13
Language
plaintext
Engine docsOpen ↗

Data flow diagram

Structured-analysis data flow diagrams in the Gane–Sarson notation: external entities, numbered processes, data stores and labelled flows, at any level (context diagram through level 2+). Trust boundaries are first-class, which makes the same figure a STRIDE threat model. Balance checking catches the classic defects — a store that is only ever read, a process with input and no output (a black hole) or output and no input (a miracle).

Templates
13
Language
plaintext
Engine docsOpen ↗

SIPOC

The Six Sigma scoping diagram — Suppliers, Inputs, Process, Outputs, Customers — as five aligned columns with the process steps as a spine across the middle. Optional `requirements:` rows attach CTQ criteria to inputs and outputs, so the chart that usually says what a process touches also says what 'good' means for each one. The first artefact of every DMAIC define phase.

Templates
13
Language
plaintext
Engine docsOpen ↗

Control chart (SPC)

Statistical process control charts — I-MR, X̄-R, X̄-S, p, np, c and u — with the centre line and control limits computed from the data by the standard unbiasing constants (d2, d3, c4, A2, A3, B3, B4, D3, D4), never typed in. All eight Nelson rules run over the plotted series and every violation is marked and listed, and process capability (Cp, Cpk, Pp, Ppk) is computed when specification limits are given. A chart whose limits were guessed is not a control chart.

Templates
13
Language
plaintext
Engine docsOpen ↗

Fault tree (FTA)

Quantitative fault-tree analysis in the IEC 61025 notation. Build the tree from AND / OR / k-out-of-n / INHIBIT gates over basic events carrying probabilities, and the engine computes the top-event probability exactly rather than by the rare-event approximation, ranks each basic event's Birnbaum and Fussell-Vesely importance, and enumerates the minimal cut sets. The safety analysis a boxes-and-lines drawing cannot perform.

Templates
13
Language
plaintext
Engine docsOpen ↗

Queueing network

Queueing networks solved, not merely drawn. Declare stations with arrival and service rates and server counts — M/M/1, M/M/c, M/M/c/K and M/G/1 are all supported — wire them with routing probabilities, and the engine solves the traffic equations and reports utilisation, mean queue length, waiting time and sojourn time per station plus the end-to-end response time, flagging any station whose utilisation reaches one as the bottleneck that makes the whole network unstable.

Templates
13
Language
plaintext
Engine docsOpen ↗

Event storming

Alberto Brandolini's event-storming board in its canonical colours: orange domain events on a timeline, blue commands, yellow actors, purple policies, pink external systems, green read models and lilac aggregates — with hotspots called out in red. Swimlanes group by bounded context, and the engine flags the shapes DDD says are suspect: an event no command produces, a command with no aggregate, an aggregate with no events.

Templates
13
Language
plaintext
Engine docsOpen ↗

Gantt & critical path

Project schedules COMPUTED from durations, dependencies and a working calendar — never placed by hand. Forward and backward pass for the critical path, total and free float drawn as a whisker on every bar, all four precedence relationships (FS, SS, FF, SF) with lead and lag, calendar-day durations for waiting periods, resources and over-allocation, baselines and deadlines. Every date written in the document is treated as a claim and checked against the computation: a stated start that its own predecessors contradict, a missed deadline, one person booked on two overlapping tasks, and a dependency loop that has no schedule at all. `view network` draws the same schedule as an activity-on-node precedence network instead of bars — one document, two views, one computation, so they cannot disagree.

Templates
13
Language
plaintext
Engine docsOpen ↗

Cloud architecture

Cloud and platform architecture across AWS, Azure, Google Cloud and Kubernetes — 331 resource types, real containment (account, region, network, zone, subnet, cluster, namespace) and typed edges. Checks the architecture as well as drawing it: a datastore reachable from the internet with no ingress on the path, a datastore in a public subnet, a tier claiming redundancy from inside one zone, an edge crossing a network boundary with no gateway.

Templates
13
Language
plaintext
Engine docsOpen ↗

Decision tree

Decision analysis with the rollback computed: decision squares, chance circles and outcome triangles, expected values rolled back from the payoffs and printed on every node, the optimal choice marked at every decision and rejected alternatives struck through. Probabilities under every chance node are validated to sum to 1 (`p rest` takes the complement), branch costs are charged on the way through, and an optional risk profile, a cumulative comparison of first choices, a stochastic-dominance check and exponential-utility certainty equivalents show what the expected value hides. Payoffs can be money, QALYs, or costs to minimise.

Templates
12
Language
plaintext
Engine docsOpen ↗

Risk matrix

A risk register and its heat map, scored rather than coloured by hand. Rate each risk on likelihood and impact (1–5 or named levels, 2–7 levels per axis), list its controls, residual and target positions, owner and actions, and the engine computes L×I scores and band ratings. It draws the matrix with risk ids in their cells and arrows from inherent to residual. It sorts the register, totals exposure and its reduction, and flags risks above appetite. Bands are configurable (thresholds, named ranges or a cell-by-cell matrix such as the ICAO tolerability matrix). The validator reports a residual worse than the inherent rating, a reduction no control explains, a risk with no owner, and a band that leaves a score unrated.

Templates
12
Language
plaintext
Engine docsOpen ↗

Customer journey map

Customer and user journey maps with the arithmetic done: stages as columns with rows for goals, actions, touchpoints, thoughts, pain points, opportunities and owners/metrics, a persona card, and an emotion curve scored −2…+2 across the stages with a face on every stage. The engine computes the moment of truth (the lowest-scored stage, ties named), the peak, the pain points per stage, the sharpest drop and the average, and flags stages that carry pain points with no opportunity against them.

Templates
12
Language
plaintext
Engine docsOpen ↗

Decision matrix (weighted / Pugh)

Weighted decision matrices and Pugh concept selection, with the arithmetic done for you. Weights are normalised and checked (two levels with groups) and scores can use any scale. Measured criteria such as cost or latency are rescaled onto that scale, and must-have knock-outs exclude an option whatever its total. You get totals, a ranking and a highlighted winner, plus the question a steering committee actually asks: how far would one weight have to move before the answer changed? It is answered for every challenger, alongside dominated options and criteria that separate nothing. Write +/S/− against a datum and the same notation becomes a Pugh matrix with Σ+, ΣS, Σ− and a weighted net.

Templates
12
Language
plaintext
Engine docsOpen ↗

Roadmap

Product, technology and programme roadmaps whose arrows are CHECKED against their dates. Lanes for themes or teams; items placed by quarter, month, half, year or on a Now / Next / Later horizon; four statuses (planned, in progress, done, at risk), each with a symbol as well as a colour so the figure survives greyscale; milestones on a band of their own with guide lines, or inside a lane; and a status date that draws the today line. Every dependency is checked: a prerequisite still running when its dependent starts is drawn red and named with the overlap in days, a loop or an unknown prerequisite is an error, and with a status date the engine reports what is overdue and anything marked done or in progress before it was due to start.

Templates
12
Language
plaintext
Engine docsOpen ↗

RACI matrix

Native RACI / RASCI responsibility-assignment matrices. Declare columns with `roles: PM, Design, Eng`, then one task per line — `Launch approval | A, I, R` — using R / A / S / C / I codes, combinations like `A/R`, and `-` for none (case-insensitive, extra or missing cells forgiven). Renders a zebra-striped matrix of colour-coded chips with a role header and a compact legend. The who-does-what chart every project kickoff rebuilds in a spreadsheet.

Templates
11
Language
plaintext
Engine docsOpen ↗

DMN

DMN 1.4 decision requirement diagrams and decision tables, with every hit policy and — the point of the engine — a real check of the table's input space: which rules can both fire, and which inputs no rule answers. A column with no declared domain is reported as undecidable rather than quietly assumed complete.

Templates
11
Language
plaintext
Engine docsOpen ↗

GRAFCET / SFC

IEC 60848 GRAFCET and IEC 61131-3 sequential function charts: numbered steps with their actions, transitions with their receptivities, and the divergences and convergences — both alternative (OR) and simultaneous (AND) — drawn with the correct single and double bars. Initial steps get their double border, and the engine checks the well-formedness rule every hand-drawn SFC violates: steps and transitions must strictly alternate.

Templates
11
Language
plaintext
Engine docsOpen ↗

Bytefield

compat: bytefield-svg

Byte-field / packet-layout diagrams. Designed for binary protocol specifications, file-format documentation, network packet structure, register maps. Tiny Clojure-flavoured DSL — `(draw-column-headers)`, `(draw-box "flags" [{:span 4}])` — renders the kind of figure you see in RFCs.

Templates
10
Language
clojure
Engine docsOpen ↗

Railroad / syntax diagram

Native railroad (syntax) diagrams for grammars, json.org style. One production per block — `rule ::= term | other` — with sequences, stacked `|` choices, `[optional]` bypasses, `{repetition}` loops and `(grouping)`. Quoted terminals render as rounded pills, bare nonterminals as rectangles, and every rule in the document renders as its own stacked diagram. The standard way language and API docs explain syntax.

Templates
10
Language
plaintext
Engine docsOpen ↗

User story map

Jeff Patton style user story maps: the activity backbone across the top, user tasks beneath it, and stories sliced into horizontal releases. Estimates roll up per release and per column, so the walking skeleton is visibly the smallest slice that spans the backbone rather than an assertion in a caption. The planning artefact a flat backlog throws away.

Templates
10
Language
plaintext
Engine docsOpen ↗

Business model canvas

The Osterwalder Business Model Canvas and Ash Maurya's Lean Canvas from one nine-block syntax — `layout: lean` switches the grid and the block names. Notes carry optional confidence marks, so the canvas shows which blocks are still guesses, and the cost/revenue strip totals any figures you put in it. The strategy one-pager every deck rebuilds by hand.

Templates
10
Language
plaintext
Engine docsOpen ↗

Spaghetti diagram

The lean motion study: place stations on a floor plan, trace the path a person, patient or part actually takes, and the engine measures it — total travel distance, per-leg distances, revisit counts and the crossings that mark a layout fighting its own flow. Multiple traced actors overlay in distinct colours so a before/after layout change is one figure, with the metres saved computed rather than claimed.

Templates
10
Language
plaintext
Engine docsOpen ↗

Ladder logic (PLC)

IEC 61131-3 ladder diagrams: rungs of normally-open and normally-closed contacts, coils, set/reset latches, timers (TON/TOF/TP) and counters (CTU/CTD), drawn between the power rails with parallel branches for OR logic. The engine evaluates each rung against a declared input state, so the drawing can show which rungs are true — the industrial control diagram nothing else in a diagram tool renders.

Templates
10
Language
plaintext
Engine docsOpen ↗

ArchiMate

ArchiMate 3.2 enterprise architecture: business, application, technology, motivation, strategy and migration layers, with all eleven relationships drawn to the specification's own line styles and end decorations. The layer is derived from the element type rather than declared beside it, so the band and the colour cannot disagree.

Templates
10
Language
plaintext
Engine docsOpen ↗

Choropleth (GIS)

Native choropleth maps for geography, demography, and epidemiology. Ships with a compact built-in atlas of world countries (ISO-3166 alpha-2) as a stylised hex-tile grid — no TopoJSON download required. Provide values per country code and the engine colour-scales them automatically. Bonus: a US states map (USPS codes) for state-level dashboards. Bigger atlas / SVG-accurate borders coming in a follow-up.

Templates
9
Language
plaintext
Engine docsOpen ↗

Music notation

Native single-voice music notation. Parses a minimal ABC-style syntax (note letters with optional accidentals, durations, bar lines) and engraves it on a treble or bass clef with key signature, time signature, and a basic note / rest set (whole, half, quarter, eighth, sixteenth). Suitable for music theory teaching, melody jotting, and lecture slides; not a full engraver — chords, beams, slurs, and complex articulations will come in a follow-up.

Templates
9
Language
plaintext
Engine docsOpen ↗

Sequence & interaction

UML sequence diagrams CHECKED against the call stack they describe, not merely drawn. Every arrow is drawn where it was typed, so what makes an interaction wrong leaves almost nothing behind in the picture: a reply returned to the wrong participant moves one arrowhead one column and changes three of the figure's ninety-five elements, and a reader who sees a reply come back reads it as a reply coming back. A message to a destroyed lifeline and a deadlock leave nothing at all. This engine decides seventeen questions the drawing cannot show: a reply returned to a participant that never made the call, a message to a lifeline after its destruction cross, a call answered on only one branch of an `alt`, a cycle of calls none of which can be answered, two concurrent branches racing to address one service, two `alt` guards that both accept the same input, and a band of inputs no branch handles. Messages state whether they are a call, a reply or a fire-and-forget, because an arrowhead does not — and a document imported from a notation that leaves that unstated is reported as unchecked rather than analysed on a guess.

Templates
9
Language
plaintext
Engine docsOpen ↗

Packet diagrams (packetdiag)

compat: packetdiag

Protocol packet layouts in a column-based DSL — write `0-15: source port` `16-31: destination port` and every bit-field renders with the correct width. Ideal for RFC-style protocol documentation; complementary to bytefield-svg.

Templates
8
Language
plaintext
Engine docsOpen ↗

Server Rack

Native data-centre rack elevations, and the only one that checks whether the rack can actually be built. Set the height in `units` and place each `device` at a top U with a height, optionally with its `power` and `weight`; the engine reports two devices claiming the same unit (drawn one over the other, so the figure shows one where the document declares two), any device that reaches past either end of the rack (silently clamped by every tool that draws what it is handed), free capacity and the tallest device that would still fit, and the declared draw and mass against the rack's stated budgets.

Templates
8
Language
plaintext
Engine docsOpen ↗

SysML v2 (structure)

Model-based systems engineering in the OMG SysML v2 textual notation — parts, ports, interfaces, item flows, specialization and redefinition — drawn as an interconnection view and a decomposition, and CHECKED. An internal block diagram hides everything that decides whether a system can actually be wired: a port's directions live in another box on another page, its conjugation is a single `~`, an interface's end order appears nowhere on the page at all. Delete one tilde from a correct fuel system and the figure is pixel-identical while every direction reverses. This engine decides thirteen questions the picture cannot show — two ports that both supply and never consume, an interface bound the wrong way round, a flow carrying a payload its destination does not accept, a redefinition that misspelt what it replaces and silently declared a second feature beside it, a composition cycle no instance of which can be built — and it distinguishes a defect in the model from a library it does not carry, so `import ISQ::*` produces gaps rather than forty false errors.

Templates
8
Language
plaintext
Engine docsOpen ↗

ERD

compat: BurntSushi/erd

Pure entity-relationship DSL. `[Person]` then `*name` for keys, `firstName` for fields, `Person *--1 Company {label: "works at"}` for relations. Renders through Graphviz internally, so layout is rock-solid even for 50-entity schemas. The cleanest text-to-ERD pipeline available.

Templates
7
Language
erd
Engine docsOpen ↗

WireViz

compat: WireViz

Cable-harness diagrams for hardware engineering. Declare `connectors` and `cables` in YAML, list the `connections` between pin/wire pairs, and WireViz produces an annotated wiring schematic with conductor colours, pin labels, and a bill-of-materials. The first-class way to document any wiring loom — robotics, automotive, audio, lab rigs.

Templates
7
Language
yaml
Engine docsOpen ↗

Floor plan (2D)

Native top-down architectural floor plans. Declare walls as coordinate pairs, drop doors (with arc) and windows (double-line), label rooms with optional areas, and place furniture from a built-in library (sofa, bed, table, fridge, stove, sink, toilet, shower, bathtub, desk, tv, plant). Renders crisp printable SVG with an optional dot grid and scale bar.

Templates
7
Language
plaintext
Engine docsOpen ↗

Floor plan (3D)

Same DSL as Floor plan (2D), rendered in axonometric / isometric SVG with extruded walls — a printable 3D view, no WebGL. Pick `view iso | front | side | top` and set the wall `height`. This is intentionally a viewer (not an editable SketchUp / Revit-style modeller); for diagram, deck, and PDF use-cases an axonometric SVG covers 80 % of the use case at 5 % of the engineering cost.

Templates
7
Language
plaintext
Engine docsOpen ↗

Floor plan (3D WebGL)

Real-time 3D viewer for floor plans, rendered with Three.js. Same DSL as the 2D / axonometric floor-plan engines — walls extrude up, rooms become 3D volumes, and you can orbit / pan / zoom with the mouse or pinch on touch. This is Phase 1 of the editable architecture modeller; phase 2 will add raycaster click-to-edit on walls, doors, and furniture so you can model interactively while the DSL stays the source of truth.

Templates
6
Language
plaintext
Engine docsOpen ↗

P&ID (chemical eng)

Native process & instrumentation diagrams for chemical engineering. Declare equipment (vessel, reactor, heat-exchanger, pump, compressor, tank, column, separator), instruments (gauges + control loops in ISA-5.1-style balloons), and piping connections between them. Supports stream tags, valves (gate, ball, check, control), and tagged process variables (PV, FV, LV, TV).

Templates
6
Language
plaintext
Engine docsOpen ↗

Evidence — the research graph

Evidence turns a body of literature into a living, citation-grounded evidence graph. Ask a research question, then find papers (type a topic or paste DOIs → OpenAlex) or paste their text. The flowss Evidence Agent extracts claims, verifies each one against its exact source quote, and assembles a concept graph — concepts as nodes, typed relationships as edges (supports / contradicts / increases / moderates…). Nothing enters the graph uncited.

  • Reads many ways: concept map, causal DAG, GRADE evidence table + Summary of Findings, PRISMA flow, sortable evidence table.
  • Finds the tension: automatic contradiction detection, research-gap finding, and leave-one-out robustness (which studies are load-bearing).
  • Goes deeper: causal back-door adjustment sets (do-calculus), mechanism path tracing, temporal trends, and quantitative effect synthesis with forest plots.
  • Ask it anything: a streaming, multi-turn flowss Evidence Agent grounded in the graph's verified claims — with clickable [n] citations — that summarizes, drafts review narrative, explains conflicts, and suggests one-click literature searches to fill the gaps.
  • Lives & exports: save a project and re-run as new papers appear (the graph diffs); export a Markdown report, BibTeX/RIS, a claims CSV, or a portable JSON graph.
flowss agents

Talking to your studio's agent

Every workspace has its own named agent — the flowss Studio Agent in the Studio, the flowss Lab Agent in Lab, the flowss Weave Agent in Weave, the flowss Evidence Agent in Evidence, and so on. A few things to know:

  1. It's mode-aware. An empty canvas triggers generate; a populated one triggers refine — the model sees the current source and edits it instead of starting over.
  2. Suggestions appear after every response. The model always returns three short next-prompts (“Add error branch”, “Switch to top-down”, “Use colour-blind safe palette”). One click to apply.
  3. Dictate it. The microphone button is wired to the Web Speech API — works in Chrome, Edge, and Safari 14+.
  4. BYO key. Paste a key from any supported provider in Settings and the daily quota goes away. By default the key stays in your browser's localStorage; if you opt in to saving it to your account, it is stored encrypted (AES-256-GCM) — never in plain text, and never logged.

Good prompts: “transformer attention block, four heads, residual + LayerNorm”. Bad prompts: “make a diagram” (we have no idea what about).

Headless API

Everything flowss does is scriptable. Create a key in your account and call the endpoints with Authorization: Bearer glyp_…. Great for CI, docs generators, and pipelines.

  • POST /api/render — render the source of any engine drawn on the server to SVG/PNG, no headless browser. Needs an API key (or a signed-in session); 60 renders a day on every plan, and Lab, BPMN and Weave engines need Starter.
  • POST /api/v1/diagram — generate a diagram from a natural-language prompt (engine, title, source, embed URL).
  • POST /api/v1/convert — convert a diagram between engines: lossless transpiler pairs run with no model, every other pair is an AI conversion charged to your key.
  • POST /api/v1/detect — identify which engine a snippet of source is written for (engine, confidence, source).
  • POST /api/v1/generate — deterministically render a structural source (SQL, DBML, D2, Graphviz, OpenAPI) to Mermaid.
  • POST /api/v1/diff — structural, meaning-aware diff of two sources (added/removed/renamed nodes & edges) for PR checks.
  • POST /api/v1/synthesize — assemble a citation-grounded Evidence graph from claims (deterministic, no AI).
  • POST /api/v1/analyze — the full Evidence analytical layer in one call: graph + contradictions/gaps + GRADE certainty, robustness, temporal trends, Summary of Findings, and a recommendations agenda.

Every endpoint also responds to GET with its own JSON schema/usage, so a CI step can call them with nothing but curl.

Exports

SVG · PNG · PDF · TikZ · source

  • SVG — vector export, perfect for slides, papers and further editing.
  • PNG (high-DPI) — choose 1×, 2× or 3× (the default) and an optional transparent background; the choice is remembered.
  • PDF — a single A4 page, oriented to fit the diagram, centred with a 10 mm margin; or every figure in one PDF.
  • Copy as PNG / SVG — straight to the clipboard, ready to paste into a document or chat.
  • TikZ / LaTeX — drop into Overleaf, opens straight as a figure.
  • Source — copy or download the engine's native source to round-trip with the workspace or version in git.
  • Share link — the diagram source is encoded directly in the URL. Share the link with someone else; they sign in (free) to open it in the editor and make changes.

Every studio's formats and options are listed on Exporting your work.

Shortcuts

Get fluent fast

Global
  • Command palette⌘K · Ctrl+K
  • Flow Dock (continue anywhere)⌘J · Ctrl+J
  • Open the flowss Studio Agent⌘E · Ctrl+E
  • Icon library⌘I · Ctrl+I
  • Ask the studio's agent (island studios)⌘/ · Ctrl+/
  • This shortcuts overlay?
Weave canvas
  • Add shapeDrag from palette
  • Move / renameDrag / double-click
  • Duplicate⌘D · Ctrl+D
  • DeleteDelete · Backspace
  • PanDrag empty canvas
  • ZoomWheel · pinch
Studio
  • New figure⌥⌘N · Ctrl+Alt+N
  • Show / hide the code⌘\ · Ctrl+\
  • Settings⌘, · Ctrl+,
  • Toggle theme⇧⌘D · Ctrl+Shift+D
  • TikZ / LaTeX export⌘L · Ctrl+L
  • Export PNG · SVG⇧⌘P · ⇧⌘S
Modals
  • Close anythingEscape
  • Send AI prompt⌘ Enter · Ctrl+Enter
  • Voice dictationMic button (Chrome / Edge / Safari)

Question? hello@flowss.ai · See live numbers on /status. Every key, studio by studio, is on Keyboard shortcuts.

© 2026 Voranox Inc. flowss — Flow Systems Studio. All rights reserved.

This documentation, its text and its examples are protected by copyright. Engine and format names are trademarks of their respective owners — see the terms and copyright and licences.