Incremental formalization
Every node has a mode: ◯ empty (a placeholder), ◎ prompt (natural-language intent), or ◉ formal (full λεξις syntax). You sketch loosely and tighten node by node — no rewrite, no tool switch.
Morphē — The Visual Canvas for System Design
Design, prototype, and formalize systems visually. Drag nodes, connect workflows, and let λεξις be the code behind every diagram.
What Makes Morphē Different
Every diagramming tool draws boxes and arrows. These four things are what a whiteboard, a Mermaid diagram, or a Figma canvas structurally cannot do — and they are shipping in the beta today.
Anchored & verified nodes
A node can be anchored to a real file or symbol in your repository, then verified against it. Green means the anchor still resolves; red is a computed gap; amber means the code moved. The diagram makes a claim you can check.
Activity lenses
Morphē reads an overlay from the Context Router and paints lens badges on each node — the real work Claude Code did on that part of the system. The map reflects what actually happened, not what you drew months ago.
Strata hand-off
Select a subgraph and Morphē compiles it to a Markdown + YAML skeleton — a precise brief that hands straight back to Pragma, so the next Claude Code session starts already oriented.
◯
Empty
A named placeholder — structure without commitment.
◎
Prompt
Intent in natural language, ready for the escape hatch.
◉
Formal
Full λεξις — typed, verifiable, executable.
Three Ways to Use Morphē
Visual Thinker
Drag and drop nodes, connect them visually. No code required. Design flowcharts, architecture diagrams, and data pipelines with a familiar canvas interface.
Human-AI Collaborator
Describe intent in natural language — the escape hatch translates it to λεξις nodes automatically. Review, refine, and iterate alongside the AI in the same canvas.
Formal Methods Engineer
Write λεξις directly. Attach proof obligations, type annotations, and formal invariants to your diagrams. The diagram is the code — no translation layer.
The Node System
Every element in Morphē is a typed node grounded in λεξις primitives. Nodes carry semantic meaning — not just visual shape. Connections encode ontological relationships: data flow, control flow, composition, dependency, and formal proof links.
The five ontological levels (L1–L5) let you move fluidly from informal sketch to formally verified specification within the same canvas. Each level adds precision without forcing premature commitment.
Escape Hatch — From Language to Diagram
Stuck on how to model something? Use the escape hatch. Type a description in plain English (or any language), and Morphē generates the corresponding λεξις node graph — ready to inspect, edit, and formalize further.
The escape hatch is powered by Claude (Anthropic). It requires either your own Anthropic API key or a Morphē proxy plan.
Embed Anywhere
Morphē ships as a standard Lit Web Component — <morphe-canvas>.
Drop it into any HTML page, React app, Vue project, or documentation site.
Zero framework lock-in.
<script type="module" src="https://cdn.jsdelivr.net/npm/@lexis/morphe-canvas"></script>
<morphe-canvas
template="blank"
theme="light"
api-key="YOUR_ANTHROPIC_KEY"
></morphe-canvas>Everything in Morphē
A visual canvas grounded in formal semantics, designed for human–AI collaboration from day one. Here’s what you get.
Visual Canvas
Node-based diagram editor
Drag, drop, and connect nodes on an infinite canvas. Pan, zoom, select, and group with standard canvas interactions. Keyboard-first design for power users.
Ten λεξις primitives
Every node is grounded in a λεξις primitive across three families: Context, State (closed); Agent, Effect, Branch, Flow, Parallel (open); Time, Proof, Fail (bordered). No arbitrary shapes — every element has semantic meaning.
Five ontological levels
Move from informal sketch (L1) to formal proof (L5) within the same diagram. Each level adds precision without forcing you to restart or change tools.
Real-time validation
Morphē highlights structural inconsistencies as you draw. Broken edges, orphaned nodes, and invalid level transitions are flagged immediately — not at export time.
Multi-diagram workspace
Work on multiple diagrams side-by-side in a tabbed workspace. Link diagrams via context frames for fractal system architecture.
Human-AI Collaboration
Escape hatch
Type a description of any system in natural language. Morphē sends it to Claude and generates the corresponding λεξις node graph — ready to inspect, edit, and formalize further. Requires an Anthropic API key.
AI-generated node icons
Each node can have an AI-generated icon that visually represents its semantic role. Icons are generated on demand and cached locally — no repeated API calls for the same concept.
Proof obligation assist
Morphē assists with writing formal invariants and proof obligations. Claude reviews your constraints and flags potential contradictions or incomplete specifications.
BYO API key
Your Anthropic API key stays local. Prompts go directly from Morphē to Anthropic — we never see your diagram content. Alternatively, use a Morphē proxy plan for managed access.
Integration & Export
Embeddable web component
Ship Morphē anywhere as a standard Lit Web Component — <morphe-canvas>. Drop it into React, Vue, plain HTML, or documentation sites. Zero framework lock-in. Available via npm and CDN.
Export formats
Export diagrams to SVG (scalable, embed in docs), PNG (1x, 2x, 4x), and .morphe JSON (lossless, version-controlled). Import/export λεξις source text for use outside the canvas.
Local file system
Diagrams are saved as .morphe JSON files — plain text, human-readable, git-friendly. No proprietary binary formats. Works with any version control system.
Link sharing
Generate a read-only link to share a diagram snapshot. Recipients view it in a browser without installing anything. No account required to view.
Part of noema.tools
Everything in noema.tools is an atom — a minimal, typed, addressable unit of meaning. Morphē atomizes software architecture.
And the tools talk to each other: your Claude Code sessions in Pragma become project memory that Morphē reads back as a live map of the real code.
See how the ecosystem fits together →