Most diagram tools are one surface with one mental model. You either write code or you drag boxes; you either think in UML or you think in flowcharts. Glyph's bet is that the surface should match the task, not the tool — so it ships several, and they all sit on the same model.
Studio — code to diagram, 69 engines
The home base. Type (or AI-generate) source in any of 69 engines — Mermaid, PlantUML, Graphviz, D2, Vega-Lite, Wardley, circuits, anatomy, music, and on — and render live. Errors trigger a one-click AI fix; labels are click-to-edit; everything exports to PNG, SVG, PDF, and TikZ. One toolbar, one co-pilot, one export pipeline, every engine.
Lab — the research-native language
Lab runs GlyphScript, our typed language for research diagrams: cohorts with sample sizes, trial arms, KPIs, project models with real numeric attributes. The canvas is editable both ways — drag a shape and the source updates at the exact byte range; edit the source and the canvas re-renders. No regex guessing.
Weave — drag-and-drop, done right
The Lucidchart experience, free: 20 shape kinds, swimlanes, container groups, motion animations, a minimap, AWS/K8s/BPMN stencils. For when you want to think with your hands, not a syntax.
Biology — figures that look designed
BioRender-style illustrated figures with 130+ curated shapes plus live search across Iconify's 200k+ icons (42k bundled offline). Click to insert, arrange, label — publication-ready biology without the subscription.
BPMN — process modeling by the spec
A dedicated BPMN 2.0 modeller: pools, lanes, gateways, events, tasks — the real notation, not an approximation.
Sketch — the hand-drawn whiteboard
Weave with the Excalidraw aesthetic pre-armed, plus an infinite canvas and handwriting OCR. For the rough, fast, thinking-out-loud diagram.
Evidence — from papers to a cited graph
And the newest: Evidence turns a body of literature into a living, cited evidence graph — contradictions, gaps, GRADE certainty, causal analysis, and a grounded research assistant. (We wrote it up separately.)
The point: one model underneath
Here's the part that matters. Whichever surface you open, the work lives in the same source format. You can move between surfaces without redoing anything, share any diagram as a permalink with the source encoded right in the URL, render any of them headlessly through the API, or embed them live in a doc. The surface is how you think today; the model is what persists.
Pick the one that fits the diagram in your head. They're all already open.